diff --git a/paper/benchmarkset.bib b/paper/benchmarkset.bib index a748e31..e701cb7 100644 --- a/paper/benchmarkset.bib +++ b/paper/benchmarkset.bib @@ -756,15 +756,18 @@ @article{merski_homologous_2015 @article{lim_sensitivity_2016, title = {Sensitivity in Binding Free Energies due to Protein Reorganization}, + volume = {12}, issn = {1549-9618, 1549-9626}, doi = {10.1021/acs.jctc.6b00532}, language = {en}, - timestamp = {2016-08-15T18:43:59Z}, + timestamp = {2016-12-08T23:08:28Z}, + number = {9}, urldate = {2016-08-15}, journal = {Journal of Chemical Theory and Computation}, author = {Lim, Nathan M. and Wang, Lingle and Abel, Robert and Mobley, David L}, month = jul, year = {2016}, + pages = {4620--4631}, file = {acs%2Ejctc%2E6b00532.pdf:/Users/dmobley/Library/Application Support/Zotero/Profiles/i2jd8b87.default/zotero/storage/SVUZNG5Q/acs%2Ejctc%2E6b00532.pdf:application/pdf} } @@ -1199,10 +1202,11 @@ @article{rocklin_blind_2013 @misc{docker_what_2015, title = {What Is {{Docker}}?}, abstract = {Docker is an open platform to build, ship and run distributed applications anywhere.}, - timestamp = {2016-09-02T18:53:12Z}, + timestamp = {2016-12-06T22:26:20Z}, urldate = {2016-09-02}, howpublished = {\url{https://www.docker.com/what-docker}}, journal = {Docker}, + author = {{Docker Inc.}}, year = {2015-05-14T16:17:40-07:00}, file = {Snapshot:/Users/dmobley/Library/Application Support/Zotero/Profiles/i2jd8b87.default/zotero/storage/3KTW9FEP/what-docker.html:text/html} } @@ -2439,9 +2443,9 @@ @article{hajduk_statistical_2008 } @misc{christ_binding_2016, - address = {Boston, MA}, title = {Binding Affinity Prediction from Molecular Simulations: {{A}} New Standard Method in Structure-Based Drug Design?}, - timestamp = {2016-09-02T16:06:32Z}, + timestamp = {2016-12-08T23:06:46Z}, + howpublished = {\url{dx.doi.org/10.7490/f1000research.1112651.1}}, author = {Christ, Clara D.}, month = may, year = {2016}, @@ -2449,9 +2453,9 @@ @misc{christ_binding_2016 } @misc{cui_affinity_2016, - address = {Boston, MA}, title = {Affinity {{Predictions}} with {{FEP}}+: {{A Different Perspective}} on {{Performance}} and {{Utility}}}, - timestamp = {2016-09-02T16:12:43Z}, + timestamp = {2016-12-08T23:06:55Z}, + howpublished = {\url{dx.doi.org/10.7490/f1000research.1112773.1}}, author = {Cui, Guanglei}, month = may, year = {2016}, @@ -2459,9 +2463,9 @@ @misc{cui_affinity_2016 } @misc{verras_free_2016, - address = {Boston, MA}, title = {Free {{Energy Perturbation}} at {{Merck}}: {{Benchmarking}} against {{Faster Methods}}}, - timestamp = {2016-09-02T16:11:50Z}, + timestamp = {2016-12-08T23:07:14Z}, + howpublished = {\url{http://www.alchemistry.org/wiki/images/c/c3/Vertex_FreeEnergyWorkshop2016_AV.pdf}}, author = {Verras, Andreas}, month = may, year = {2016}, @@ -3024,11 +3028,31 @@ @article{yin_overview_2016 file = {Full Text PDF:/Users/dmobley/Library/Application Support/Zotero/Profiles/i2jd8b87.default/zotero/storage/22989I64/Yin et al. - 2016 - Overview of the SAMPL5 host–guest challenge Are w.pdf:application/pdf;Snapshot:/Users/dmobley/Library/Application Support/Zotero/Profiles/i2jd8b87.default/zotero/storage/5NTTWBRW/s10822-016-9974-4.html:text/html} } +@article{Liu:2007:Nucl.AcidsRes., + title = {{{BindingDB}}: A Web-Accessible Database of Experimentally Determined Protein\textendash{}ligand Binding Affinities}, + volume = {35}, + issn = {0305-1048, 1362-4962}, + shorttitle = {{{BindingDB}}}, + doi = {10.1093/nar/gkl999}, + abstract = {BindingDB (http://www.bindingdb.org) is a publicly accessible database currently containing $\sim$20 000 experimentally determined binding affinities of protein\textendash{}ligand complexes, for 110 protein targets including isoforms and mutational variants, and $\sim$11 000 small molecule ligands. The data are extracted from the scientific literature, data collection focusing on proteins that are drug-targets or candidate drug-targets and for which structural data are present in the Protein Data Bank. The BindingDB website supports a range of query types, including searches by chemical structure, substructure and similarity; protein sequence; ligand and protein names; affinity ranges and molecular weight. Data sets generated by BindingDB queries can be downloaded in the form of annotated SDfiles for further analysis, or used as the basis for virtual screening of a compound database uploaded by the user. The data in BindingDB are linked both to structural data in the PDB via PDB IDs and chemical and sequence searches, and to the literature in PubMed via PubMed IDs.}, + language = {en}, + timestamp = {2016-09-27T23:25:45Z}, + number = {suppl 1}, + urldate = {2016-09-27}, + journal = {Nucl. Acids Res.}, + author = {Liu, Tiqing and Lin, Yuhmei and Wen, Xin and Jorissen, Robert N. and Gilson, Michael K.}, + month = jan, + year = {2007}, + pages = {D198--D201}, + file = {Full Text PDF:/Users/dmobley/Library/Application Support/Zotero/Profiles/i2jd8b87.default/zotero/storage/ZCD57MX2/Liu et al. - 2007 - BindingDB a web-accessible database of experiment.pdf:application/pdf;Snapshot:/Users/dmobley/Library/Application Support/Zotero/Profiles/i2jd8b87.default/zotero/storage/ATP2T8EM/D198.html:text/html}, + pmid = {17145705} +} + @misc{Abel:2016:vertex, - address = {Boston, MA}, title = {Accelerating Drug Discovery with Free Energy Calculations}, - timestamp = {2016-11-17T19:43:12Z}, + timestamp = {2016-12-06T22:17:31Z}, urldate = {2016-11-17}, + howpublished = {\url{http://www.alchemistry.org/wiki/images/e/eb/Vertex_talk_5_15_2016_clean3.pdf}}, author = {Abel, Robert}, month = may, year = {2016}, @@ -3054,10 +3078,830 @@ @article{Gathiaka:2016:JComputAidedMolDes file = {Full Text PDF:/Users/dmobley/Library/Application Support/Zotero/Profiles/i2jd8b87.default/zotero/storage/XD77JTXE/Gathiaka et al. - 2016 - D3R grand challenge 2015 Evaluation of protein–li.pdf:application/pdf;Snapshot:/Users/dmobley/Library/Application Support/Zotero/Profiles/i2jd8b87.default/zotero/storage/I6GRFPMA/s10822-016-9946-8.html:text/html} } +@article{Marquez:2004:J.Am.Chem.Soc., + title = {Mechanism of {{Host}}-{{Guest Complexation}} by {{Cucurbituril}}}, + volume = {126}, + issn = {0002-7863}, + doi = {10.1021/ja0319846}, + abstract = {The factors affecting host-guest complexation between the molecular container compound cucurbit[6]uril (CB6) and various guests in aqueous solution are studied, and a detailed complexation mechanism in the presence of cations is derived. The formation of the supramolecular complex is studied in detail for cyclohexylmethylammonium ion as guest. The kinetics and thermodynamics of complexation is monitored by NMR as a function of temperature, salt concentration, and cation size. The binding constants and the ingression rate constants decrease with increasing salt concentration and cation-binding constant, in agreement with a competitive binding of the ammonium site of the guest and the metal cation with the ureido carbonyl portals of CB6. Studies as a function of guest size indicate that the effective container volume of the CB6 cavity is approximately 105 \AA{}3. It is suggested that larger guests are excluded for two reasons:\, a high activation barrier for ingression imposed by the tight CB6 portals and a destabilization of the complex due to steric repulsion inside. For example, in the case of the nearly spherical azoalkane homologues 2,3-diazabicyclo[2.2.1]hept-2-ene (DBH, volume ca. 96 \AA{}3) and 2,3-diazabicyclo[2.2.2]oct-2-ene (DBO, volume ca. 110 \AA{}3), the former forms the CB6 complex promptly with a sizable binding constant (1300 M-1), while the latter does not form a complex even after several months at optimized complexation conditions. Molecular mechanics calculations are performed for several CB6/guest complexes. A qualitative agreement is found between experimental and calculated activation energies for ingression as a function of both guest size and state of protonation. The potential role of constrictive binding by CB6 is discussed.}, + timestamp = {2016-12-06T22:40:06Z}, + number = {18}, + urldate = {2016-12-06}, + journal = {J. Am. Chem. Soc.}, + author = {M{\'a}rquez, C{\'e}sar and Hudgins, Robert R. and Nau, Werner M.}, + month = may, + year = {2004}, + pages = {5806--5816}, + file = {ACS Full Text PDF w/ Links:/Users/dmobley/Library/Application Support/Zotero/Profiles/i2jd8b87.default/zotero/storage/ERRPBQ7D/Márquez et al. - 2004 - Mechanism of Host−Guest Complexation by Cucurbitur.pdf:application/pdf;ACS Full Text Snapshot:/Users/dmobley/Library/Application Support/Zotero/Profiles/i2jd8b87.default/zotero/storage/RHMAE6CS/ja0319846.html:text/html} +} + +@article{Fitzgerald:1995:ProteinScience, + title = {The Role of Aspartate-235 in the Binding of Cations to an Artificial Cavity at the Radical Site of Cytochrome c Peroxidase}, + volume = {4}, + issn = {1469-896X}, + doi = {10.1002/pro.5560040919}, + abstract = {The activated state of cytochrome c peroxidase, compound ES, contains a cation radical on the Trp-191 side chain. We recently reported that replacing this tryptophan with glycine creates a buried cavity at the active site that contains ordered solvent and that will specifically bind substituted imidazoles in their protonated cationic forms (Fitzgerald MM, Churchill MJ, McRee DE, Goodin DB, 1994, Biochemistry 55:3807\textendash{}3818). Proposals that a nearby carboxylate, Asp-235, and competing monovalent cations should modulate the affinity of the W191G cavity for ligand binding are addressed in this study. Competitive binding titrations of the imidazolium ion to W191G as a function of [K+] show that potassium competes weakly with the binding of imidazoles. The dissociation constant observed for potassium binding (18 mM) is more than 3, 000-fold higher than that for 1, 2-dimethylimidazole (5.5 $\mu$M) in the absence of competing cations. Significantly, the W191G-D235N double mutant shows no evidence for binding imidazoles in their cationic or neutral forms, even though the structure of the cavity remains largely unperturbed by replacement of the carboxylate. Refined crystallographic 5-values of solvent positions indicate that the weakly bound potassium in W191G is significantly depopulated in the double mutant. These results demonstrate that the buried negative charge of Asp-235 is an essential feature of the cation binding determinant and indicate that this carboxylate plays a critical role in stabilizing the formation of the Trp-191 radical cation.}, + language = {en}, + timestamp = {2016-12-06T23:28:26Z}, + number = {9}, + urldate = {2016-12-06}, + journal = {Protein Science}, + author = {Fitzgerald, Melissa M. and Trester, Michelle L. and Jensen, Gerard M. and Mcree, Duncan E. and Goodin, David B.}, + month = sep, + year = {1995}, + keywords = {cytochrome c peroxidase,engineered binding site,enzyme-substrate binding,protein cavities,Protein engineering,tryptophan radical cation,X-ray crystallography}, + pages = {1844--1850}, + file = {Full Text PDF:/Users/dmobley/Library/Application Support/Zotero/Profiles/i2jd8b87.default/zotero/storage/3PME9DKM/Fitzgerald et al. - 1995 - The role of aspartate-235 in the binding of cation.pdf:application/pdf;Snapshot:/Users/dmobley/Library/Application Support/Zotero/Profiles/i2jd8b87.default/zotero/storage/AI9B64S9/abstract.html:text/html} +} + +@article{Bento:2014:NuclAcidsRes, + title = {The {{ChEMBL}} Bioactivity Database: An Update}, + volume = {42}, + issn = {0305-1048}, + shorttitle = {The {{ChEMBL}} Bioactivity Database}, + doi = {10.1093/nar/gkt1031}, + timestamp = {2016-12-06T23:38:45Z}, + number = {D1}, + urldate = {2016-12-06}, + journal = {Nucl Acids Res}, + author = {Bento, A. Patr{\'\i}cia and Gaulton, Anna and Hersey, Anne and Bellis, Louisa J. and Chambers, Jon and Davies, Mark and Kr{\"u}ger, Felix A. and Light, Yvonne and Mak, Lora and McGlinchey, Shaun and Nowotka, Michal and Papadatos, George and Santos, Rita and Overington, John P.}, + month = jan, + year = {2014}, + pages = {D1083--D1090}, + file = {Full Text PDF:/Users/dmobley/Library/Application Support/Zotero/Profiles/i2jd8b87.default/zotero/storage/V65U5F7B/Bento et al. - 2014 - The ChEMBL bioactivity database an update.pdf:application/pdf;Snapshot:/Users/dmobley/Library/Application Support/Zotero/Profiles/i2jd8b87.default/zotero/storage/KF5FE9TU/gkt1031.html:text/html} +} + +@article{Kim:2016:NucleicAcidsRes., + title = {{{PubChem Substance}} and {{Compound}} Databases}, + volume = {44}, + issn = {1362-4962}, + doi = {10.1093/nar/gkv951}, + abstract = {PubChem (https://pubchem.ncbi.nlm.nih.gov) is a public repository for information on chemical substances and their biological activities, launched in 2004 as a component of the Molecular Libraries Roadmap Initiatives of the US National Institutes of Health (NIH). For the past 11 years, PubChem has grown to a sizable system, serving as a chemical information resource for the scientific research community. PubChem consists of three inter-linked databases, Substance, Compound and BioAssay. The Substance database contains chemical information deposited by individual data contributors to PubChem, and the Compound database stores unique chemical structures extracted from the Substance database. Biological activity data of chemical substances tested in assay experiments are contained in the BioAssay database. This paper provides an overview of the PubChem Substance and Compound databases, including data sources and contents, data organization, data submission using PubChem Upload, chemical structure standardization, web-based interfaces for textual and non-textual searches, and programmatic access. It also gives a brief description of PubChem3D, a resource derived from theoretical three-dimensional structures of compounds in PubChem, as well as PubChemRDF, Resource Description Framework (RDF)-formatted PubChem data for data sharing, analysis and integration with information contained in other databases.}, + language = {eng}, + timestamp = {2016-12-06T23:39:33Z}, + number = {D1}, + journal = {Nucleic Acids Res.}, + author = {Kim, Sunghwan and Thiessen, Paul A. and Bolton, Evan E. and Chen, Jie and Fu, Gang and Gindulyte, Asta and Han, Lianyi and He, Jane and He, Siqian and Shoemaker, Benjamin A. and Wang, Jiyao and Yu, Bo and Zhang, Jian and Bryant, Stephen H.}, + month = jan, + year = {2016}, + keywords = {Databases; Chemical,Internet,Molecular Structure,Pharmaceutical Preparations,Software}, + pages = {D1202--1213}, + pmid = {26400175}, + pmcid = {PMC4702940} +} + +@article{StefanicAnderluh:2005:J.Med.Chem., + title = {Toward a {{Novel Class}} of {{Antithrombotic Compounds}} with {{Dual Function}}. {{Discovery}} of 1,4-{{Benzoxazin}}-3({{4H}})-One {{Derivatives Possessing Thrombin Inhibitory}} and {{Fibrinogen Receptor Antagonistic Activities}}}, + volume = {48}, + issn = {0022-2623}, + doi = {10.1021/jm048984g}, + abstract = {A novel class of potential antithrombotic compounds with moderate thrombin inhibitory and fibrinogen receptor antagonistic activity is described. Combination of anticoagulant and antiaggregatory activity in the same molecular entity is presented as a new promising approach in the search for novel antithrombotic agents.}, + timestamp = {2016-12-07T00:04:18Z}, + number = {9}, + urldate = {2016-12-07}, + journal = {J. Med. Chem.}, + author = {{\v S}tefani{\v c} Anderluh, Petra and Anderluh, Marko and Ila{\v s}, Janez and Mravljak, Janez and Sollner Dolenc, Marija and Stegnar, Mojca and Kikelj, Danijel}, + month = may, + year = {2005}, + pages = {3110--3113}, + file = {ACS Full Text PDF w/ Links:/Users/dmobley/Library/Application Support/Zotero/Profiles/i2jd8b87.default/zotero/storage/S6BUP62K/Štefanič Anderluh et al. - 2005 - Toward a Novel Class of Antithrombotic Compounds w.pdf:application/pdf;ACS Full Text Snapshot:/Users/dmobley/Library/Application Support/Zotero/Profiles/i2jd8b87.default/zotero/storage/7JQ327CK/jm048984g.html:text/html} +} + +@article{Ueno:2005:Bioorganic&MedicinalChemistryLetters, + title = {Discovery of Novel Tetrahydroisoquinoline Derivatives as Potent and Selective Factor {{Xa}} Inhibitors}, + volume = {15}, + issn = {0960-894X}, + doi = {10.1016/j.bmcl.2004.10.033}, + abstract = {A series of novel 2,7-disubstituted tetrahydroisoquinoline derivatives were designed and synthesized. Among these derivatives, compounds 1 and 2 (JTV-803) exhibited potent inhibitory activity against FXa and good selectivity with respect to other serine proteases (thrombin, plasmin, and trypsin). In addition, compound 2 exhibited potent anti-FXa activity after intravenous and oral administration to cynomolgus monkey, and showed a dose-dependent antithrombotic effect in a rat model of venous thrombosis.}, + timestamp = {2016-12-07T00:04:47Z}, + number = {1}, + urldate = {2016-12-07}, + journal = {Bioorganic \& Medicinal Chemistry Letters}, + author = {Ueno, Hiroshi and Yokota, Katsuyuki and Hoshi, Jun-ichi and Yasue, Katsutaka and Hayashi, Mikio and Uchida, Itsuo and Aisaka, Kazuo and Hase, Yasunori and Katoh, Susumu and Cho, Hidetsura}, + month = jan, + year = {2005}, + keywords = {Antithrombotic effect on venous thrombosis in rats,Inhibition of human FXa,JTV-803,Potent and selective factor Xa (FXa) inhibitor,Tetrahydroisoquinoline derivatives}, + pages = {185--189}, + file = {ScienceDirect Full Text PDF:/Users/dmobley/Library/Application Support/Zotero/Profiles/i2jd8b87.default/zotero/storage/9AXWI3GN/Ueno et al. - 2005 - Discovery of novel tetrahydroisoquinoline derivati.pdf:application/pdf;ScienceDirect Snapshot:/Users/dmobley/Library/Application Support/Zotero/Profiles/i2jd8b87.default/zotero/storage/8KI7KDDG/S0960894X04012612.html:text/html} +} + +@article{Putta:2005:J.Med.Chem., + title = {Conformation {{Mining}}:\, {{An Algorithm}} for {{Finding Biologically Relevant Conformations}}}, + volume = {48}, + issn = {0022-2623}, + shorttitle = {Conformation {{Mining}}}, + doi = {10.1021/jm049066l}, + abstract = {Discovering essential features shared by active compounds, an important step in drug-design, is complicated by conformational flexibility. We present a new algorithm to efficiently mine the conformational space of multiple actives and find small subsets of conformations likely to be biologically relevant. The approach identifies chemical and steric similarities between actives, providing insight into features important for binding when structural data are absent. Validation studies (thrombin and CDK2 data) produce alignments similar to protein-based alignments.}, + timestamp = {2016-12-07T00:05:18Z}, + number = {9}, + urldate = {2016-12-07}, + journal = {J. Med. Chem.}, + author = {Putta, Santosh and Landrum, Gregory A. and Penzotti, Julie E.}, + month = may, + year = {2005}, + pages = {3313--3318}, + file = {ACS Full Text PDF w/ Links:/Users/dmobley/Library/Application Support/Zotero/Profiles/i2jd8b87.default/zotero/storage/E62XJRMF/Putta et al. - 2005 - Conformation Mining An Algorithm for Finding Bio.pdf:application/pdf;ACS Full Text Snapshot:/Users/dmobley/Library/Application Support/Zotero/Profiles/i2jd8b87.default/zotero/storage/KT8M2CT5/jm049066l.html:text/html} +} + +@article{Dullweber:2001:JournalofMolecularBiology, + title = {Factorising Ligand Affinity: A Combined Thermodynamic and Crystallographic Study of Trypsin and Thrombin inhibition\textdagger\textdagger{}2}, + volume = {313}, + issn = {0022-2836}, + shorttitle = {Factorising Ligand Affinity}, + doi = {10.1006/jmbi.2001.5062}, + abstract = {The binding of a series of low molecular weight ligands towards trypsin and thrombin has been studied by isothermal titration calorimetry and protein crystallography. In a series of congeneric ligands, surprising changes of protonation states occur and are overlaid on the binding process. They result from induced pKa shifts depending on the local environment experienced by the ligand and protein functional groups in the complex (induced dielectric fit). They involve additional heat effects that must be corrected before any conclusion on the binding enthalpy ($\Delta$H) and entropy ($\Delta$S) can be drawn. After correction, trends in both contributions can be interpreted in structural terms with respect to the hydrogen bond inventory or residual ligand motions. For all inhibitors studied, a strong negative heat capacity change ($\Delta$Cp) is detected, thus binding becomes more exothermic and entropically less favourable with increasing temperature. Due to a mutual compensation, Gibbs free energy remains virtually unchanged. The strong negative $\Delta$Cp value cannot solely be explained by the removal of hydrophobic surface portions of the protein or ligand from water exposure. Additional contributions must be considered, presumably arising from modulations of the local water structure, changes in vibrational modes or other ordering parameters. For thrombin, smaller negative $\Delta$Cp values are observed for ligand binding in the presence of sodium ions compared to the other alkali ions, probably due to stabilising effects on the protein or changes in the bound water structure.}, + timestamp = {2016-12-07T00:06:29Z}, + number = {3}, + urldate = {2016-12-07}, + journal = {Journal of Molecular Biology}, + author = {Dullweber, Frank and Stubbs, Milton T and Musil, \DJ{}or\dj{}e and St{\"u}rzebecher, J{\"o}rg and Klebe, Gerhard}, + month = oct, + year = {2001}, + keywords = {drug design,enthalpy-entropy compensation,isothermal titration calorimetry,Thermodynamics,X-ray crystallography}, + pages = {593--614}, + file = {ScienceDirect Full Text PDF:/Users/dmobley/Library/Application Support/Zotero/Profiles/i2jd8b87.default/zotero/storage/P5T8DHJC/Dullweber et al. - 2001 - Factorising ligand affinity a combined thermodyna.pdf:application/pdf;ScienceDirect Snapshot:/Users/dmobley/Library/Application Support/Zotero/Profiles/i2jd8b87.default/zotero/storage/K9JNP53T/S0022283601950624.html:text/html} +} + +@article{Filippakopoulos:2010:Nature, + title = {Selective Inhibition of {{BET}} Bromodomains}, + volume = {468}, + copyright = {\textcopyright{} 2010 Nature Publishing Group, a division of Macmillan Publishers Limited. All Rights Reserved.}, + issn = {0028-0836}, + doi = {10.1038/nature09504}, + abstract = {Epigenetic proteins are intently pursued targets in ligand discovery. So far, successful efforts have been limited to chromatin modifying enzymes, or so-called epigenetic `writers' and `erasers'. Potent inhibitors of histone binding modules have not yet been described. Here we report a cell-permeable small molecule (JQ1) that binds competitively to acetyl-lysine recognition motifs, or bromodomains. High potency and specificity towards a subset of human bromodomains is explained by co-crystal structures with bromodomain and extra-terminal (BET) family member BRD4, revealing excellent shape complementarity with the acetyl-lysine binding cavity. Recurrent translocation of BRD4 is observed in a genetically-defined, incurable subtype of human squamous carcinoma. Competitive binding by JQ1 displaces the BRD4 fusion oncoprotein from chromatin, prompting squamous differentiation and specific antiproliferative effects in BRD4-dependent cell lines and patient-derived xenograft models. These data establish proof-of-concept for targeting protein\textendash{}protein interactions of epigenetic `readers', and provide a versatile chemical scaffold for the development of chemical probes more broadly throughout the bromodomain family.}, + language = {en}, + timestamp = {2016-12-07T00:24:17Z}, + number = {7327}, + urldate = {2016-12-07}, + journal = {Nature}, + author = {Filippakopoulos, Panagis and Qi, Jun and Picaud, Sarah and Shen, Yao and Smith, William B. and Fedorov, Oleg and Morse, Elizabeth M. and Keates, Tracey and Hickman, Tyler T. and Felletar, Ildiko and Philpott, Martin and Munro, Shonagh and McKeown, Michael R. and Wang, Yuchuan and Christie, Amanda L. and West, Nathan and Cameron, Michael J. and Schwartz, Brian and Heightman, Tom D. and La Thangue, Nicholas and French, Christopher A. and Wiest, Olaf and Kung, Andrew L. and Knapp, Stefan and Bradner, James E.}, + month = dec, + year = {2010}, + keywords = {Cancer,Drug discovery,Structural biology}, + pages = {1067--1073}, + file = {Full Text PDF:/Users/dmobley/Library/Application Support/Zotero/Profiles/i2jd8b87.default/zotero/storage/MEWH4ECC/Filippakopoulos et al. - 2010 - Selective inhibition of BET bromodomains.pdf:application/pdf;Snapshot:/Users/dmobley/Library/Application Support/Zotero/Profiles/i2jd8b87.default/zotero/storage/DC38J6EN/nature09504.html:text/html} +} + +@article{Chung:2011:J.Med.Chem., + title = {Discovery and {{Characterization}} of {{Small Molecule Inhibitors}} of the {{BET Family Bromodomains}}}, + volume = {54}, + issn = {0022-2623}, + doi = {10.1021/jm200108t}, + abstract = {Epigenetic mechanisms of gene regulation have a profound role in normal development and disease processes. An integral part of this mechanism occurs through lysine acetylation of histone tails which are recognized by bromodomains. While the biological and structural characterization of many bromodomain containing proteins has advanced considerably, the therapeutic tractability of this protein family is only now becoming understood. This paper describes the discovery and molecular characterization of potent (nM) small molecule inhibitors that disrupt the function of the BET family of bromodomains (Brd2, Brd3, and Brd4). By using a combination of phenotypic screening, chemoproteomics, and biophysical studies, we have discovered that the protein\textendash{}protein interactions between bromodomains and acetylated histones can be antagonized by selective small molecules that bind at the acetylated lysine recognition pocket. X-ray crystal structures of compounds bound into bromodomains of Brd2 and Brd4 elucidate the molecular interactions of binding and explain the precisely defined stereochemistry required for activity.}, + timestamp = {2016-12-07T00:25:04Z}, + number = {11}, + urldate = {2016-12-07}, + journal = {J. Med. Chem.}, + author = {Chung, Chun-wa and Coste, Herv{\'e} and White, Julia H. and Mirguet, Olivier and Wilde, Jonathan and Gosmini, Romain L. and Delves, Chris and Magny, Sylvie M. and Woodward, Robert and Hughes, Stephen A. and Boursier, Eric V. and Flynn, Helen and Bouillot, Anne M. and Bamborough, Paul and Brusq, Jean-Marie G. and Gellibert, Fran{\c c}oise J. and Jones, Emma J. and Riou, Alizon M. and Homes, Paul and Martin, Sandrine L. and Uings, Iain J. and Toum, J{\'e}r{\^o}me and Cl{\'e}ment, Catherine A. and Boullay, Anne-B{\'e}n{\'e}dicte and Grimley, Rachel L. and Blandel, Florence M. and Prinjha, Rab K. and Lee, Kevin and Kirilovsky, Jorge and Nicodeme, Edwige}, + month = jun, + year = {2011}, + pages = {3827--3838}, + file = {ACS Full Text PDF w/ Links:/Users/dmobley/Library/Application Support/Zotero/Profiles/i2jd8b87.default/zotero/storage/XBIJVDS5/Chung et al. - 2011 - Discovery and Characterization of Small Molecule I.pdf:application/pdf;ACS Full Text Snapshot:/Users/dmobley/Library/Application Support/Zotero/Profiles/i2jd8b87.default/zotero/storage/WS78V7BX/jm200108t.html:text/html} +} + +@article{Hewings:2012:J.Med.Chem., + title = {Progress in the {{Development}} and {{Application}} of {{Small Molecule Inhibitors}} of {{Bromodomain}}\textendash{}{{Acetyl}}-Lysine {{Interactions}}}, + volume = {55}, + issn = {0022-2623}, + doi = {10.1021/jm300915b}, + abstract = {Bromodomains, protein modules that recognize and bind to acetylated lysine, are emerging as important components of cellular machinery. These acetyl-lysine (KAc) ``reader'' domains are part of the write\textendash{}read\textendash{}erase concept that has been linked with the transfer of epigenetic information. By reading KAc marks on histones, bromodomains mediate protein\textendash{}protein interactions between a diverse array of partners. There has been intense activity in developing potent and selective small molecule probes that disrupt the interaction between a given bromodomain and KAc. Rapid success has been achieved with the BET family of bromodomains, and a number of potent and selective probes have been reported. These compounds have enabled linking of the BET bromodomains with diseases, including cancer and inflammation, suggesting that bromodomains are druggable targets. Herein, we review the biology of the bromodomains and discuss the SAR for the existing small molecule probes. The biology that has been enabled by these compounds is summarized.}, + timestamp = {2016-12-07T00:26:11Z}, + number = {22}, + urldate = {2016-12-07}, + journal = {J. Med. Chem.}, + author = {Hewings, David S. and Rooney, Timothy P. C. and Jennings, Laura E. and Hay, Duncan A. and Schofield, Christopher J. and Brennan, Paul E. and Knapp, Stefan and Conway, Stuart J.}, + month = nov, + year = {2012}, + pages = {9393--9413}, + file = {ACS Full Text PDF w/ Links:/Users/dmobley/Library/Application Support/Zotero/Profiles/i2jd8b87.default/zotero/storage/A48MQ565/Hewings et al. - 2012 - Progress in the Development and Application of Sma.pdf:application/pdf;ACS Full Text Snapshot:/Users/dmobley/Library/Application Support/Zotero/Profiles/i2jd8b87.default/zotero/storage/8RXR4QQS/jm300915b.html:text/html} +} + +@article{Bamborough:2015:J.Med.Chem., + title = {Structure-{{Based Optimization}} of {{Naphthyridones}} into {{Potent ATAD2 Bromodomain Inhibitors}}}, + volume = {58}, + issn = {0022-2623}, + doi = {10.1021/acs.jmedchem.5b00773}, + abstract = {ATAD2 is a bromodomain-containing protein whose overexpression is linked to poor outcomes in a number of different cancer types. To date, no potent and selective inhibitors of the bromodomain have been reported. This article describes the structure-based optimization of a series of naphthyridones from micromolar leads with no selectivity over the BET bromodomains to inhibitors with sub-100 nM ATAD2 potency and 100-fold BET selectivity.}, + timestamp = {2016-12-07T00:27:14Z}, + number = {15}, + urldate = {2016-12-07}, + journal = {J. Med. Chem.}, + author = {Bamborough, Paul and Chung, Chun-wa and Furze, Rebecca C. and Grandi, Paola and Michon, Anne-Marie and Sheppard, Robert J. and Barnett, Heather and Diallo, Hawa and Dixon, David P. and Douault, Clement and Jones, Emma J. and Karamshi, Bhumika and Mitchell, Darren J. and Prinjha, Rab K. and Rau, Christina and Watson, Robert J. and Werner, Thilo and Demont, Emmanuel H.}, + month = aug, + year = {2015}, + pages = {6151--6178}, + file = {ACS Full Text PDF w/ Links:/Users/dmobley/Library/Application Support/Zotero/Profiles/i2jd8b87.default/zotero/storage/FDSGPG34/Bamborough et al. - 2015 - Structure-Based Optimization of Naphthyridones int.pdf:application/pdf;ACS Full Text Snapshot:/Users/dmobley/Library/Application Support/Zotero/Profiles/i2jd8b87.default/zotero/storage/XDBZFXET/acs.jmedchem.html:text/html} +} + +@article{Ran:2015:J.Med.Chem., + title = {Structure-{{Based Design}} of $\gamma$-{{Carboline Analogues}} as {{Potent}} and {{Specific BET Bromodomain Inhibitors}}}, + volume = {58}, + issn = {0022-2623}, + doi = {10.1021/acs.jmedchem.5b00613}, + abstract = {Small-molecule inhibitors of bromodomain and extra terminal proteins (BET), including BRD2, BRD3, and BRD4 proteins have therapeutic potential for the treatment of human cancers and other diseases and conditions. In this paper, we report the design, synthesis, and evaluation of $\gamma$-carboline-containing compounds as a new class of small-molecule BET inhibitors. The most potent inhibitor (compound 18, RX-37) obtained from this study binds to BET bromodomain proteins (BRD2, BRD3, and BRD4) with Ki values of 3.2\textendash{}24.7 nM and demonstrates high selectivity over other non-BET bromodomain-containing proteins. Compound 18 potently and selectively inhibits cell growth in human acute leukemia cell lines harboring the rearranged mixed lineage leukemia 1 gene. We have determined a cocrystal structure of 18 in complex with BRD4 BD2 at 1.4 \AA{} resolution, which provides a solid structural basis for the compound's high binding affinity and for its further structure-based optimization. Compound 18 represents a promising lead compound for the development of a new class of therapeutics for the treatment of human cancer and other conditions.}, + timestamp = {2016-12-07T00:28:37Z}, + number = {12}, + urldate = {2016-12-07}, + journal = {J. Med. Chem.}, + author = {Ran, Xu and Zhao, Yujun and Liu, Liu and Bai, Longchuan and Yang, Chao-Yie and Zhou, Bing and Meagher, Jennifer L. and Chinnaswamy, Krishnapriya and Stuckey, Jeanne A. and Wang, Shaomeng}, + month = jun, + year = {2015}, + pages = {4927--4939}, + file = {ACS Full Text PDF w/ Links:/Users/dmobley/Library/Application Support/Zotero/Profiles/i2jd8b87.default/zotero/storage/BFKU5W5U/Ran et al. - 2015 - Structure-Based Design of γ-Carboline Analogues as.pdf:application/pdf;ACS Full Text Snapshot:/Users/dmobley/Library/Application Support/Zotero/Profiles/i2jd8b87.default/zotero/storage/RIFC3PVD/acs.jmedchem.html:text/html} +} + +@article{Zhang:2013:J.Med.Chem., + title = {Structure-{{Guided Design}} of {{Potent Diazobenzene Inhibitors}} for the {{BET Bromodomains}}}, + volume = {56}, + issn = {0022-2623}, + doi = {10.1021/jm401334s}, + abstract = {BRD4, characterized by two acetyl-lysine binding bromodomains and an extra-terminal (ET) domain, is a key chromatin organizer that directs gene activation in chromatin through transcription factor recruitment, enhancer assembly, and pause release of the RNA polymerase II complex for transcription elongation. BRD4 has been recently validated as a new epigenetic drug target for cancer and inflammation. Our current knowledge of the functional differences of the two bromodomains of BRD4, however, is limited and is hindered by the lack of selective inhibitors. Here, we report our structure-guided development of diazobenzene-based small-molecule inhibitors for the BRD4 bromodomains that have over 90\% sequence identity at the acetyl-lysine binding site. Our lead compound, MS436, through a set of water-mediated interactions, exhibits low nanomolar affinity (estimated Ki of 30\textendash{}50 nM), with preference for the first bromodomain over the second. We demonstrated that MS436 effectively inhibits BRD4 activity in NF-$\kappa$B-directed production of nitric oxide and proinflammatory cytokine interleukin-6 in murine macrophages. MS436 represents a new class of bromodomain inhibitors and will facilitate further investigation of the biological functions of the two bromodomains of BRD4 in gene expression.}, + timestamp = {2016-12-07T00:29:09Z}, + number = {22}, + urldate = {2016-12-07}, + journal = {J. Med. Chem.}, + author = {Zhang, Guangtao and Plotnikov, Alexander N. and Rusinova, Elena and Shen, Tong and Morohashi, Keita and Joshua, Jennifer and Zeng, Lei and Mujtaba, Shiraz and Ohlmeyer, Michael and Zhou, Ming-Ming}, + month = nov, + year = {2013}, + pages = {9251--9264}, + file = {ACS Full Text PDF w/ Links:/Users/dmobley/Library/Application Support/Zotero/Profiles/i2jd8b87.default/zotero/storage/W7QEXHSC/Zhang et al. - 2013 - Structure-Guided Design of Potent Diazobenzene Inh.pdf:application/pdf;ACS Full Text Snapshot:/Users/dmobley/Library/Application Support/Zotero/Profiles/i2jd8b87.default/zotero/storage/G883ANFM/jm401334s.html:text/html} +} + +@article{Gosmini:2014:J.Med.Chem., + title = {The {{Discovery}} of {{I}}-{{BET726}} ({{GSK1324726A}}), a {{Potent Tetrahydroquinoline ApoA1 Up}}-{{Regulator}} and {{Selective BET Bromodomain Inhibitor}}}, + volume = {57}, + issn = {0022-2623}, + doi = {10.1021/jm5010539}, + abstract = {Through their function as epigenetic readers of the histone code, the BET family of bromodomain-containing proteins regulate expression of multiple genes of therapeutic relevance, including those involved in tumor cell growth and inflammation. BET bromodomain inhibitors have profound antiproliferative and anti-inflammatory effects which translate into efficacy in oncology and inflammation models, and the first compounds have now progressed into clinical trials. The exciting biology of the BETs has led to great interest in the discovery of novel inhibitor classes. Here we describe the identification of a novel tetrahydroquinoline series through up-regulation of apolipoprotein A1 and the optimization into potent compounds active in murine models of septic shock and neuroblastoma. At the molecular level, these effects are produced by inhibition of BET bromodomains. X-ray crystallography reveals the interactions explaining the structure\textendash{}activity relationships of binding. The resulting lead molecule, I-BET726, represents a new, potent, and selective class of tetrahydroquinoline-based BET inhibitors.}, + timestamp = {2016-12-07T00:30:20Z}, + number = {19}, + urldate = {2016-12-07}, + journal = {J. Med. Chem.}, + author = {Gosmini, Romain and Nguyen, Van Loc and Toum, J{\'e}r{\^o}me and Simon, Christophe and Brusq, Jean-Marie G. and Krysa, Gael and Mirguet, Olivier and Riou-Eymard, Alizon M. and Boursier, Eric V. and Trottet, Lionel and Bamborough, Paul and Clark, Hugh and Chung, Chun-wa and Cutler, Leanne and Demont, Emmanuel H. and Kaur, Rejbinder and Lewis, Antonia J. and Schilling, Mark B. and Soden, Peter E. and Taylor, Simon and Walker, Ann L. and Walker, Matthew D. and Prinjha, Rab K. and Nicod{\`e}me, Edwige}, + month = oct, + year = {2014}, + pages = {8111--8131}, + file = {ACS Full Text PDF w/ Links:/Users/dmobley/Library/Application Support/Zotero/Profiles/i2jd8b87.default/zotero/storage/A9FR87ID/Gosmini et al. - 2014 - The Discovery of I-BET726 (GSK1324726A), a Potent .pdf:application/pdf;ACS Full Text Snapshot:/Users/dmobley/Library/Application Support/Zotero/Profiles/i2jd8b87.default/zotero/storage/TP56CTVE/jm5010539.html:text/html} +} + +@article{Yang:2016:BioorganicChemistry, + title = {Design, Synthesis and Biological Evaluation of Dihydroquinoxalinone Derivatives as {{BRD4}} Inhibitors}, + volume = {68}, + issn = {0045-2068}, + doi = {10.1016/j.bioorg.2016.08.009}, + abstract = {BRD4 plays a key role in transcriptional regulation. Recent biological and pharmacological studies have demonstrated that bromodomain-containing protein 4 (BRD4) is a viable drug target for cancer treatment. In this study, we synthesized a series of dihydroquinoxalinone derivatives and evaluated their BRD4 inhibitory activities, obtaining compound 5i with IC50 value of 73 nM of binding activity in BRD4(1) and 258 nM of cellular activity in MV-4-11 cancer cell lines. Docking studies were performed to explain the structure-activity relationship. Based on its potent biochemical and anti-proliferative activity, the novel BRD4 inhibitor compound 5i, is a promising lead compound for further investigation.}, + timestamp = {2016-12-07T00:30:50Z}, + urldate = {2016-12-07}, + journal = {Bioorganic Chemistry}, + author = {Yang, Yifei and Zhao, Leilei and Xu, Bin and Yang, LingYun and Zhang, Jian and Zhang, Huibin and Zhou, Jinpei}, + month = oct, + year = {2016}, + keywords = {BRD4 inhibitors,Cancer,Dihydroquinoxalinone derivatives,molecular docking}, + pages = {236--244}, + file = {ScienceDirect Full Text PDF:/Users/dmobley/Library/Application Support/Zotero/Profiles/i2jd8b87.default/zotero/storage/2EES7456/Yang et al. - 2016 - Design, synthesis and biological evaluation of dih.pdf:application/pdf;ScienceDirect Snapshot:/Users/dmobley/Library/Application Support/Zotero/Profiles/i2jd8b87.default/zotero/storage/GE5FHHXD/S0045206816300827.html:text/html} +} + +@misc{Kurtzer:2016:singularity, + title = {Singularity 2.1.2 - {{Linux}} Application and Environment Containers for Science}, + timestamp = {2016-12-07T00:39:03Z}, + author = {Kurtzer, G. M.}, + month = dec, + year = {2016} +} + +@article{Connors:1997:Chem.Rev., + title = {The {{Stability}} of {{Cyclodextrin Complexes}} in {{Solution}}}, + volume = {97}, + issn = {0009-2665}, + doi = {10.1021/cr960371r}, + timestamp = {2016-12-08T22:06:24Z}, + number = {5}, + urldate = {2016-12-08}, + journal = {Chem. Rev.}, + author = {Connors, Kenneth A.}, + month = aug, + year = {1997}, + pages = {1325--1358}, + file = {ACS Full Text PDF w/ Links:/Users/dmobley/Library/Application Support/Zotero/Profiles/i2jd8b87.default/zotero/storage/QEA4A98F/Connors - 1997 - The Stability of Cyclodextrin Complexes in Solutio.pdf:application/pdf;ACS Full Text Snapshot:/Users/dmobley/Library/Application Support/Zotero/Profiles/i2jd8b87.default/zotero/storage/5PIM3A7H/cr960371r.html:text/html} +} + +@article{Carrazana:2005:J.Phys.Chem.B, + title = {Complexation of {{Adamantyl Compounds}} by $\beta$-{{Cyclodextrin}} and {{Monoaminoderivatives}}}, + volume = {109}, + issn = {1520-6106}, + doi = {10.1021/jp0505781}, + abstract = {Since the $\beta$-cyclodextrin cavity is not a smooth cone but has constrictions in the neighborhoods of the H3 and H5 atoms, the hypothesis that bulky hydrophobic guests can form two isomeric inclusion complexes (one of them, cp, is formed by the entrance of the guest by the primary side of the cavity, and the other one, cs, results from the entrance by the secondary side) is checked. Thus, the inclusion processes of two 1-substituted adamantyl derivatives (rimantidine and adamantylmethanol) with $\beta$-cyclodextrin and its two monoamino derivatives at positions 6 (6-NH2$\beta$-CD) and 3 (3-NH2$\beta$-CD) were studied. From rotating-frame Overhauser enhancement spectroscopy experiments, it was deduced that both guests form cs complexes with $\beta$-CD and 6-NH2$\beta$-CD but cp complexes with 3-NH2$\beta$-CD. In all cases, the hydrophilic group attached to the adamantyl residue protrudes toward the bulk solvent outside the cyclodextrin cavity. The thermodynamic parameters (free energy, equilibrium constant, enthalpy, and entropy) associated with the inclusion phenomena were measured by isothermal titration calorimetry experiments. From these results, the difference in the free energy for the formation of the two complexes, cs and cp, for the same host/guest system has been estimated as being 11.5 $\pm$ 0.8 kJ mol-1. This large difference explains why under normal experimental conditions only one of the two complexes (cs) is detected. It is also concluded that a hyperboloid of revolution can be a better schematic picture to represent the actual geometry of the cyclodextrin cavities than the usual smooth cone or trapezium.}, + timestamp = {2016-12-08T22:06:31Z}, + number = {19}, + urldate = {2016-12-08}, + journal = {J. Phys. Chem. B}, + author = {Carrazana, Jorge and Jover, Aida and Meijide, Francisco and Soto, Victor H. and V{\'a}zquez Tato, Jos{\'e}}, + month = may, + year = {2005}, + pages = {9719--9726}, + file = {ACS Full Text PDF w/ Links:/Users/dmobley/Library/Application Support/Zotero/Profiles/i2jd8b87.default/zotero/storage/8AGEIHZC/Carrazana et al. - 2005 - Complexation of Adamantyl Compounds by β-Cyclodext.pdf:application/pdf;ACS Full Text Snapshot:/Users/dmobley/Library/Application Support/Zotero/Profiles/i2jd8b87.default/zotero/storage/TJT8XS5I/jp0505781.html:text/html} +} + +@article{Cotner:1998:J.Org.Chem., + title = {Phosphotyrosine {{Binding}} by {{Ammonium}}- and {{Guanidinium}}-{{Modified Cyclodextrins}}}, + volume = {63}, + issn = {0022-3263}, + doi = {10.1021/jo971979i}, + timestamp = {2016-12-08T22:17:43Z}, + number = {5}, + urldate = {2016-12-08}, + journal = {J. Org. Chem.}, + author = {Cotner, Edward S. and Smith, Paul J.}, + month = mar, + year = {1998}, + pages = {1737--1739}, + file = {ACS Full Text PDF w/ Links:/Users/dmobley/Library/Application Support/Zotero/Profiles/i2jd8b87.default/zotero/storage/7E2K9T98/Cotner and Smith - 1998 - Phosphotyrosine Binding by Ammonium- and Guanidini.pdf:application/pdf;ACS Full Text Snapshot:/Users/dmobley/Library/Application Support/Zotero/Profiles/i2jd8b87.default/zotero/storage/JBFM8CSZ/jo971979i.html:text/html} +} + +@article{Wszelaka-Rylik:2013:JThermAnalCalorim, + title = {Isothermal Titration Calorimetry ({{ITC}}) Study of Natural Cyclodextrins Inclusion Complexes with Drugs}, + volume = {111}, + issn = {1388-6150, 1572-8943}, + doi = {10.1007/s10973-012-2251-4}, + abstract = {Isothermal titration calorimetry (ITC) was used to characterize inclusion complex formation of natural cyclodextrins ($\alpha$- and $\beta$-cyclodextrin) with three drugs ((+)brompheniramine, ($\pm$)brompheniramine, cyclopentolate) in aqueous solutions. ITC measurements were carried out at 298.15 K on a Microcal OMEGA ultrasensitive titration calorimeter (MicroCal Inc.). The experimental data were analyzed on the basis of the model of a single set of identical sites (ITC tutorial guide). $\beta$-CD forms inclusion complexes of stoichiometry 1:1 with the all investigated drugs. In turn, smaller molecule of $\alpha$-CD forms inclusion complexes of two different stoichiometry: with bigger molecules ((+)brompheniramine and ($\pm$)brompheniramine) of a stoichiometry 2:1 and with smaller molecules (cyclopentolate) of a stoichiometry 1:2. Based on the experimental values of equilibrium constant (K) and enthalpy of complex formation ($\Delta$H), the Gibbs energy of complex formation ($\Delta$G), and the entropy of complex formation ($\Delta$S), have been calculated, for all the investigated systems. Obtained results showed that complex formation of $\beta$-CD (bigger molecule with wider cavity compared to $\beta$-CD) with both (+)brompheniramine, ($\pm$)brompheniramine, and cyclopentolate is enthalpy driven while complexes of $\alpha$-CD with the all investigated drugs are enthalpy-entropy stabilized. This indicated that the difference in the cavity dimensions is reflecting in different driving forces of complex formation and binding modes what resulted in different stoichiometry of the obtained inclusion complexes.}, + language = {en}, + timestamp = {2016-12-08T22:17:51Z}, + number = {3}, + urldate = {2016-12-08}, + journal = {J Therm Anal Calorim}, + author = {Wszelaka-Rylik, Ma\l{}gorzata and Gierycz, Pawe\l{}}, + month = mar, + year = {2013}, + pages = {2029--2035}, + file = {Full Text PDF:/Users/dmobley/Library/Application Support/Zotero/Profiles/i2jd8b87.default/zotero/storage/82ZDS7G8/Wszelaka-Rylik and Gierycz - 2013 - Isothermal titration calorimetry (ITC) study of na.pdf:application/pdf;Snapshot:/Users/dmobley/Library/Application Support/Zotero/Profiles/i2jd8b87.default/zotero/storage/4N6SWK3Z/s10973-012-2251-4.html:text/html} +} + +@article{Shu:2007:BritishJournalofPharmacology, + title = {Cyclodextrins Sequester Neuroactive Steroids and Differentiate Mechanisms That Rate Limit Steroid Actions}, + volume = {150}, + issn = {1476-5381}, + doi = {10.1038/sj.bjp.0706973}, + abstract = {Background and purpose: +Neuroactive steroids are potent modulators of GABAA receptors and are thus of interest for their sedative, anxiolytic, anticonvulsant and anaesthetic properties. Cyclodextrins may be useful tools to manipulate neuroactive effects of steroids on GABAA receptors because cyclodextrins form inclusion complexes with at least some steroids that are active at the GABAA receptor, such as (3$\alpha$,5$\alpha$)-3-hydroxypregnan-20-one (3$\alpha$5$\alpha$P, allopregnanolone). +Experimental approach: +To assess the versatility of cyclodextrins as steroid modulators, we investigated interactions between $\gamma$-cyclodextrin and neuroactive steroids of different structural classes. +Key results: +Both a bioassay based on electrophysiological assessment of GABAA receptor function and optical measurements of cellular accumulation of a fluorescent steroid analogue suggest that $\gamma$-cyclodextrin sequesters steroids rather than directly influencing GABAA receptor function. Neither a 5$\beta$-reduced A/B ring fusion nor a sulphate group at carbon 3 affected the presumed inclusion complex formation between steroid and $\gamma$-cyclodextrin. Apparent dissociation constants for interactions between natural steroids and $\gamma$-cyclodexrin ranged from 10-60 $\mu$M. Although $\gamma$-cyclodextrin accommodates a range of natural and synthetic steroids, C11 substitutions reduced inclusion complex formation. Using $\gamma$-cyclodextrin to remove steroid not directly bound to GABAA receptors, we found that cellular retention of receptor-unbound steroid rate limits potentiation by 3$\alpha$- hydroxysteroids but not inhibition by sulphated steroids. +Conclusions and implications: +We conclude that $\gamma$-cyclodextrins can be useful, albeit non-specific, tools for terminating the actions of multiple classes of naturally occurring neuroactive steroids.British Journal of Pharmacology (2007) 150, 164\textendash{}175. doi:10.1038/sj.bjp.0706973}, + language = {en}, + timestamp = {2016-12-08T22:18:02Z}, + number = {2}, + urldate = {2016-12-08}, + journal = {British Journal of Pharmacology}, + author = {Shu, H-J and Zeng, C-M and Wang, C and Covey, D F and Zorumski, C F and Mennerick, S}, + month = jan, + year = {2007}, + keywords = {GABAA receptors,inhibitory postsynaptic currents,neurosteroid,γ-aminobutyric acid}, + pages = {164--175}, + file = {Full Text PDF:/Users/dmobley/Library/Application Support/Zotero/Profiles/i2jd8b87.default/zotero/storage/CF69GZCA/Shu et al. - 2007 - Cyclodextrins sequester neuroactive steroids and d.pdf:application/pdf;Snapshot:/Users/dmobley/Library/Application Support/Zotero/Profiles/i2jd8b87.default/zotero/storage/5EBUCFWX/abstract.html:text/html} +} + +@article{Rodriguez-Perez:2006:J.Pharm.Sci., + title = {Sertaconazole/Hydroxypropyl-$\beta$-Cyclodextrin Complexation: {{Isothermal}} Titration Calorimetry and Solubility Approaches}, + volume = {95}, + issn = {1520-6017}, + shorttitle = {Sertaconazole/Hydroxypropyl-$\beta$-Cyclodextrin Complexation}, + doi = {10.1002/jps.20661}, + abstract = {Complexation of sertaconazole (SN) with hydroxypropyl-$\beta$-cyclodextrin (HP-$\beta$-CD) was characterized by phase-solubility diagram measurements and isothermal calorimetry (ITC) in aqueous medium, and by differential scanning calorimetry (DSC), Raman spectroscopy and X-ray diffractometry in solid state. The strongest interaction was observed at pH 1.2, at which two different 1:1 complexes can be formed depending on the hydrophobic ring of the drug involved in the process. At pH 5.8 and 7.4 the likelihood of 1:2 stoichiometry increases as a consequence of the simultaneous complexation of the nonprotonized imidazolyl and the dichlorophenyl groups. In the presence of 20\% HP-$\beta$-CD, SN solubility is enhanced by a factor of 116, 107, and 5 at pH 1.2, 5.8, and 7.4, respectively. Complexation enthalpy recorded by ITC showed the same tendency which confirms the practical interest of this technique for fast screening of the potential of CDs as drug solubilizers. Solubility and dissolution rate of the drug from compacts prepared with freeze-dried complexes were significantly greater than those obtained with SN powder or compacts made with physical blends. \textcopyright{} 2006 Wiley-Liss, Inc. and the American Pharmacists Association J Pharm Sci 95: 1751\textendash{}1762, 2006}, + language = {en}, + timestamp = {2016-12-08T22:18:17Z}, + number = {8}, + urldate = {2016-12-08}, + journal = {J. Pharm. Sci.}, + author = {Rodriguez-Perez, Ana I. and Rodriguez-Tenreiro, Carmen and Alvarez-Lorenzo, Carmen and Taboada, Pablo and Concheiro, Angel and Torres-Labandeira, Juan J.}, + month = aug, + year = {2006}, + keywords = {complexation,cyclodextrins,inclusion compounds,isothermal calorimetry (ITC),sertaconazole,solubility}, + pages = {1751--1762}, + file = {Full Text PDF:/Users/dmobley/Library/Application Support/Zotero/Profiles/i2jd8b87.default/zotero/storage/U9PSP8AE/Rodriguez-Perez et al. - 2006 - Sertaconazolehydroxypropyl-β-cyclodextrin complex.pdf:application/pdf;Snapshot:/Users/dmobley/Library/Application Support/Zotero/Profiles/i2jd8b87.default/zotero/storage/2KAWPFAX/abstract.html:text/html} +} + +@article{Szente:1999:AdvancedDrugDeliveryReviews, + series = {Cyclodextrins in Drug Delivery System}, + title = {Highly Soluble Cyclodextrin Derivatives: Chemistry, Properties, and Trends in Development}, + volume = {36}, + issn = {0169-409X}, + shorttitle = {Highly Soluble Cyclodextrin Derivatives}, + doi = {10.1016/S0169-409X(98)00092-1}, + abstract = {As the first pharmaceutical products which contain highly soluble cyclodextrin (CD) derivatives (e.g. Sporanox\texttrademark=itraconazole/HP-$\beta$-CD by Janssen and Clorocil\texttrademark=chloramphenicol/methyl-$\beta$-CD by Oftalder) are already on the market it seems to be timely to give an overview on the technological and commercial aspects of the chemically modified water-soluble CDs as drug carriers. This chapter deals with the chemistry and general properties of water-soluble CDs and follows the trends in their development. The quality requirements of industrially relevant water-soluble CDs together with the quality-assurance-related analytical techniques, are thought help understand how difficult the characterization and approval processes of such novel excipients are. Literature data taken from Cyclolab's databank support the validity of statements in evaluation of trends of development of CD derivatives as pharmaceutical excipients.}, + timestamp = {2016-12-08T22:18:41Z}, + number = {1}, + urldate = {2016-12-08}, + journal = {Advanced Drug Delivery Reviews}, + author = {Szente, Lajos and Szejtli, J{\'o}zsef}, + month = mar, + year = {1999}, + keywords = {Analytical methods,CaptisolTM,Commercial aspects,Manufacturing,MolecusolTM,Quality control,Solubilization,Synthesis,Toxicity,Water-soluble CD-derivatives}, + pages = {17--28}, + file = {ScienceDirect Full Text PDF:/Users/dmobley/Library/Application Support/Zotero/Profiles/i2jd8b87.default/zotero/storage/99V2R56G/Szente and Szejtli - 1999 - Highly soluble cyclodextrin derivatives chemistry.pdf:application/pdf;ScienceDirect Snapshot:/Users/dmobley/Library/Application Support/Zotero/Profiles/i2jd8b87.default/zotero/storage/AGV4WXIB/S0169409X98000921.html:text/html} +} + +@article{Qu:2002:JournalofInclusionPhenomena, + title = {Sulfoalkyl {{Ether}} $\beta$-{{Cyclodextrin Derivatives}}: {{Synthesis}} and {{Characterizations}}}, + volume = {43}, + issn = {0923-0750, 1573-1111}, + shorttitle = {Sulfoalkyl {{Ether}} $\beta$-{{Cyclodextrin Derivatives}}}, + doi = {10.1023/A:1021255314835}, + abstract = {A series of sulfoalkyl ether $\beta$-cyclodextrin derivatives, including sulfoethyl, sulfopropyl and sulfobutyl ethyl $\beta$-cyclodextrins, have been synthesized and characterized. Each sulfoalkyl ether $\beta$-cyclodextrin is a mixture of various degrees of substitution and different positional isotherms. Elemental analysis, 1H NMR, MS, and Differential Scanning Calorimetry analysis were used to determine the average degree of substitution for each $\beta$-cyclodextrin derivative. The average degrees of substitution are 3.4, 1.6 and 2.5 for sulfoethyl, sulfopropyl and sulfobutyl ether $\beta$-cyclodextrin, respectively. The water solubility of these derivatives is substantially higher than that of $\beta$-cyclodextrin. 1H NMR indicates thatsulfoethyl ether $\beta$-cyclodextrin may have major substitution on the secondaryhydroxyl group while the major substitution in sulfopropyl and sulfobutyl ether$\beta$-cyclodextrin could be on the primary hydroxyl group. MS spectra showthat no more than one substitution occurred on a single glucose unit.}, + language = {en}, + timestamp = {2016-12-08T22:18:59Z}, + number = {3-4}, + urldate = {2016-12-08}, + journal = {Journal of Inclusion Phenomena}, + author = {Qu, Q. I. and Tucker, Edward and Christian, Sherril D.}, + month = aug, + year = {2002}, + pages = {213--222}, + file = {Full Text PDF:/Users/dmobley/Library/Application Support/Zotero/Profiles/i2jd8b87.default/zotero/storage/799B8W5E/Qu et al. - 2002 - Sulfoalkyl Ether β-Cyclodextrin Derivatives Synth.pdf:application/pdf;Snapshot:/Users/dmobley/Library/Application Support/Zotero/Profiles/i2jd8b87.default/zotero/storage/HHBVDE6W/A1021255314835.html:text/html} +} + +@article{Jindrich:2005:J.Org.Chem., + title = {Simple {{Preparation}} of {{3I}}-{{O}}-{{Substituted}} $\beta$-{{Cyclodextrin Derivatives Using Cinnamyl Bromide}}}, + volume = {70}, + issn = {0022-3263}, + doi = {10.1021/jo051339c}, + abstract = {A new method for preparation of 3I-O-substituted $\beta$-cyclodextrin derivatives was developed. Cinnamyl bromide reacts with $\beta$-cyclodextrin to form predominantly the 3I-O-cinnamyl derivative (30\% isolated yield, $>$90\% regioselectivity). After protection of the remaining cyclodextrin hydroxyls by acetylation, the cinnamyl group can be easily transformed to many other groups (exemplified by transformation to 3I-O-carboxymethyl derivative). Substitution pattern in singly modified CDs was unambiguously determined by a combination of 2D NMR techniques.}, + timestamp = {2016-12-08T22:19:13Z}, + number = {22}, + urldate = {2016-12-08}, + journal = {J. Org. Chem.}, + author = {{Jind{\v r}ich} and Ti{\v s}lerov{\'a}, Iva}, + month = oct, + year = {2005}, + pages = {9054--9055}, + file = {ACS Full Text PDF w/ Links:/Users/dmobley/Library/Application Support/Zotero/Profiles/i2jd8b87.default/zotero/storage/KNWRQZB4/Jindřich and Tišlerová - 2005 - Simple Preparation of 3I-O-Substituted β-Cyclodext.pdf:application/pdf;ACS Full Text Snapshot:/Users/dmobley/Library/Application Support/Zotero/Profiles/i2jd8b87.default/zotero/storage/CTP82WJJ/jo051339c.html:text/html} +} + +@article{Faugeras:2012:Eur.J.Org.Chem., + title = {When {{Cyclodextrins Meet Click Chemistry}}}, + volume = {2012}, + issn = {1099-0690}, + doi = {10.1002/ejoc.201200013}, + abstract = {Cyclodextrins are important building blocks in organic chemistry. This review deals with the role of click chemistry in this family of cyclic oligosaccharides, focusing on the different areas of chemistry, including chromatography, biological applications, the elaboration of superstructures, and metal detection, that benefit from this reaction. In this paper, attention is given to organic modifications by using functionalizations such as azidation and propargylation and to click chemistry grafting onto the two faces of cyclodextrins. Research papers where cyclodextrins are not directly involved in a click chemistry reaction are not considered.}, + language = {en}, + timestamp = {2016-12-08T22:19:25Z}, + number = {22}, + urldate = {2016-12-08}, + journal = {Eur. J. Org. Chem.}, + author = {Faugeras, Pierre-Antoine and Bo{\"e}ns, Benjamin and Elchinger, Pierre-Henri and Brouillette, Fran{\c c}ois and Montplaisir, Daniel and Zerrouki, Rachida and Lucas, Romain}, + month = aug, + year = {2012}, + keywords = {Click chemistry,Cycloaddition,cyclodextrins,Oligosaccharides}, + pages = {4087--4105}, + file = {Full Text PDF:/Users/dmobley/Library/Application Support/Zotero/Profiles/i2jd8b87.default/zotero/storage/ZWWBAAJD/Faugeras et al. - 2012 - When Cyclodextrins Meet Click Chemistry.pdf:application/pdf;Snapshot:/Users/dmobley/Library/Application Support/Zotero/Profiles/i2jd8b87.default/zotero/storage/NUAQXU9Q/abstract.html:text/html} +} + +@article{Mark:1994:J.Am.Chem.Soc., + title = {Convergence {{Properties}} of {{Free Energy Calculations}}: .alpha.-{{Cyclodextrin Complexes}} as a {{Case Study}}}, + volume = {116}, + issn = {0002-7863}, + shorttitle = {Convergence {{Properties}} of {{Free Energy Calculations}}}, + doi = {10.1021/ja00093a032}, + timestamp = {2016-12-08T22:19:28Z}, + number = {14}, + urldate = {2016-12-08}, + journal = {J. Am. Chem. Soc.}, + author = {Mark, Alan E. and {van Helden}, Steven P. and Smith, Paul E. and Janssen, Lambert H. M. and {van Gunsteren}, Wilfred F.}, + month = jul, + year = {1994}, + pages = {6293--6302}, + file = {ACS Full Text PDF:/Users/dmobley/Library/Application Support/Zotero/Profiles/i2jd8b87.default/zotero/storage/KSAJ4VTP/Mark et al. - 1994 - Convergence Properties of Free Energy Calculations.pdf:application/pdf;ACS Full Text Snapshot:/Users/dmobley/Library/Application Support/Zotero/Profiles/i2jd8b87.default/zotero/storage/ERNAB9KQ/ja00093a032.html:text/html} +} + +@article{Luzhkov:1999:ChemicalPhysicsLetters, + title = {Free-Energy Perturbation Calculations of Binding and Transition-State Energies: Hydrolysis of Phenyl Esters by $\beta$-Cyclodextrin}, + volume = {302}, + issn = {0009-2614}, + shorttitle = {Free-Energy Perturbation Calculations of Binding and Transition-State Energies}, + doi = {10.1016/S0009-2614(99)00109-8}, + abstract = {The use of free-energy perturbation simulations for calculating binding and transition-state energies is examined. Reactions of three phenyl esters with $\beta$-cyclodextrin are considered. The binding free energies are calculated using conventional mutation of the substrate force field from one chemical structure to another (two-state problem) in water and in the neutral inclusion complex. For calculation of activation free-energy differences between two substrates the transition states are represented by linear combinations of reactant (anionic inclusion complexes) and product (tetrahedral intermediate) force fields (four-state problem). The coefficients of this linear combination are obtained from empirical valence bond simulations of a reference substrate. The calculated relative binding and activation energies are in a good agreement with experimental data. The approximations underlying this procedure are discussed.}, + timestamp = {2016-12-08T22:19:36Z}, + number = {3\textendash{}4}, + urldate = {2016-12-08}, + journal = {Chemical Physics Letters}, + author = {Luzhkov, Victor and \AA{}qvist, Johan}, + month = mar, + year = {1999}, + pages = {267--272}, + file = {ScienceDirect Full Text PDF:/Users/dmobley/Library/Application Support/Zotero/Profiles/i2jd8b87.default/zotero/storage/DSMRSRHD/Luzhkov and Åqvist - 1999 - Free-energy perturbation calculations of binding a.pdf:application/pdf;ScienceDirect Snapshot:/Users/dmobley/Library/Application Support/Zotero/Profiles/i2jd8b87.default/zotero/storage/I24G6K6H/S0009261499001098.html:text/html} +} + +@article{Bea:2002:TheorChemAcc, + title = {Molecular Recognition by $\beta$-Cyclodextrin Derivatives: Molecular Dynamics, Free-Energy Perturbation and Molecular Mechanics/ {{Poisson}}\textendash{}{{Boltzmann}} Surface Area Goals and Problems}, + volume = {108}, + issn = {1432-881X, 1432-2234}, + shorttitle = {Molecular Recognition by $\beta$-Cyclodextrin Derivatives}, + doi = {10.1007/s00214-002-0384-4}, + abstract = {. The complexation of p-tert-butylphenyl p-tert-butylbenzoate and N-(p-tert-butylphenyl)-p-tert-butylbenzamide with a $\beta$-cyclodextrin derivative formed by two cyclodextrin units linked by a disulfide bridge on one of the C6 atoms has been studied by computational methods. The better amide solubility and the better internal interactions of the ester complex explain the experimentally observed better association constant for the ester. The free-energy perturbation methodology and molecular mechanics/Poisson\textendash{}Boltzmann surface area analysis have been used to explain the problem and to compare the results.}, + language = {en}, + timestamp = {2016-12-08T22:19:47Z}, + number = {5}, + urldate = {2016-12-08}, + journal = {Theor Chem Acc}, + author = {Be{\`a}, Ivan and Jaime, Carlos and Kollman, Peter}, + month = nov, + year = {2002}, + pages = {286--292}, + file = {Full Text PDF:/Users/dmobley/Library/Application Support/Zotero/Profiles/i2jd8b87.default/zotero/storage/GW5NE76N/Beà et al. - 2002 - Molecular recognition by β-cyclodextrin derivative.pdf:application/pdf;Snapshot:/Users/dmobley/Library/Application Support/Zotero/Profiles/i2jd8b87.default/zotero/storage/7B88QIIR/s00214-002-0384-4.html:text/html} +} + +@article{Chen:2004:BiophysicalJournal, + title = {Calculation of {{Cyclodextrin Binding Affinities}}: {{Energy}}, {{Entropy}}, and {{Implications}} for {{Drug Design}}}, + volume = {87}, + issn = {0006-3495}, + shorttitle = {Calculation of {{Cyclodextrin Binding Affinities}}}, + doi = {10.1529/biophysj.104.049494}, + abstract = {The second generation Mining Minima method yields binding affinities accurate to within 0.8 kcal/mol for the associations of $\alpha$-, $\beta$-, and $\gamma$-cyclodextrin with benzene, resorcinol, flurbiprofen, naproxen, and nabumetone. These calculations require hours to a day on a commodity computer. The calculations also indicate that the changes in configurational entropy upon binding oppose association by as much as 24 kcal/mol and result primarily from a narrowing of energy wells in the bound versus the free state, rather than from a drop in the number of distinct low-energy conformations on binding. Also, the configurational entropy is found to vary substantially among the bound conformations of a given cyclodextrin-guest complex. This result suggests that the configurational entropy must be accounted for to reliably rank docked conformations in both host-guest and ligand-protein complexes. In close analogy with the common experimental observation of entropy-enthalpy compensation, the computed entropy changes show a near-linear relationship with the changes in mean potential plus solvation energy.}, + timestamp = {2016-12-08T22:19:52Z}, + number = {5}, + urldate = {2016-12-08}, + journal = {Biophysical Journal}, + author = {Chen, Wei and Chang, Chia-En and Gilson, Michael K.}, + month = nov, + year = {2004}, + pages = {3035--3049}, + file = {ScienceDirect Full Text PDF:/Users/dmobley/Library/Application Support/Zotero/Profiles/i2jd8b87.default/zotero/storage/XPRD2IUW/Chen et al. - 2004 - Calculation of Cyclodextrin Binding Affinities En.pdf:application/pdf;ScienceDirect Snapshot:/Users/dmobley/Library/Application Support/Zotero/Profiles/i2jd8b87.default/zotero/storage/GDFMR7SZ/S0006349504737753.html:text/html} +} + +@article{Sellner:2008:J.Phys.Chem.B, + title = {Molecular {{Dynamics Simulations}} of $\beta$-{{Cyclodextrin}}-{{Aziadamantane Complexes}} in {{Water}}}, + volume = {112}, + issn = {1520-6106}, + doi = {10.1021/jp075493+}, + abstract = {Force-field-based atomistic simulations of host-guest supramolecular complexes between $\beta$-cyclodextrin and several aziadamantane derivatives have been analyzed with respect to relative orientation and interaction energies, explicitly considering solvent (water) molecules. For each case, the calculations revealed two stable orientations of the guest within the host that are different in interaction energy. Fluctuation of and correlation between characteristic properties were analyzed. Among other things, it turned out that orientation angle and inclusion depth are clearly correlated. In addition, for the unsubstituted aziadamantane, the enthalpy of complex formation was calculated and compared to experimental results.}, + timestamp = {2016-12-08T22:20:01Z}, + number = {3}, + urldate = {2016-12-08}, + journal = {J. Phys. Chem. B}, + author = {Sellner, Bernhard and Zifferer, Gerhard and Kornherr, Andreas and Krois, Daniel and Brinker, Udo H.}, + month = jan, + year = {2008}, + pages = {710--714}, + file = {ACS Full Text PDF w/ Links:/Users/dmobley/Library/Application Support/Zotero/Profiles/i2jd8b87.default/zotero/storage/5VJWZ8RB/Sellner et al. - 2008 - Molecular Dynamics Simulations of β-Cyclodextrin−A.pdf:application/pdf;ACS Full Text Snapshot:/Users/dmobley/Library/Application Support/Zotero/Profiles/i2jd8b87.default/zotero/storage/DDKANF2B/jp075493+.html:text/html} +} + +@article{Cai:2009:J.Phys.Chem.B, + title = {Inclusion {{Mechanism}} of {{Steroid Drugs}} into $\beta$-{{Cyclodextrins}}. {{Insights}} from {{Free Energy Calculations}}}, + volume = {113}, + issn = {1520-6106}, + doi = {10.1021/jp901825w}, + abstract = {The inclusion of hydrocortisone, progesterone, and testosterone into the cavity of $\beta$-cyclodextrin ($\beta$-CD) following two possible orientations was investigated using molecular dynamics simulations and free-energy calculations. The free-energy profiles that delineate the inclusion process were determined using an adaptive biasing force. The present results reveal that although the free-energy surfaces feature two local minima corresponding to a partial and a complete inclusion, the former mode is markedly preferred, irrespective of the orientation. Ranking the propensity of the three steroidal molecules to associate with $\beta$-CD, viz. progesterone $>$ testosterone $>$ hydrocortisone, is shown to be in excellent agreement with experiment. This conclusion is further supported by independent calculations relying on alchemical transformations in conjunction with free energy perturbation, wherein the relative binding free energy for the three steroids was estimated. In addition, decomposition of the potentials of mean force into free-energy contributions and significant decrease in the total hydrophobic surface area suggest that by and large, van der Waals and hydrophobic interactions constitute the main driving forces responsible for the formation of the inclusion complexes. Analysis of their structural features from the molecular dynamics trajectories brings to light different hydrogen-bonding patterns that are characterized by distinct dynamics and stabilities.}, + timestamp = {2016-12-08T22:20:06Z}, + number = {22}, + urldate = {2016-12-08}, + journal = {J. Phys. Chem. B}, + author = {Cai, Wensheng and Sun, Tingting and Liu, Peng and Chipot, Christophe and Shao, Xueguang}, + month = jun, + year = {2009}, + pages = {7836--7843}, + file = {ACS Full Text Snapshot:/Users/dmobley/Library/Application Support/Zotero/Profiles/i2jd8b87.default/zotero/storage/NMM4PFK5/jp901825w.html:text/html} +} + +@article{ShengCai:2011:CurrentOrganicChemistry, + title = {Free {{Energy Calculations}} for {{Cyclodextrin Inclusion Complexes}}}, + volume = {15}, + doi = {10.2174/138527211794518853}, + abstract = {In recent years, a variety of computational methods has been employed to investigate cyclodextrin (CD)-based systems. Among these methods, free-energy calculations occupy a prominent position, because they provide the key thermodynamic quantity that underlies the formation and the stability of CD complexes, together with the atomic-level detail often inaccessible to experiment. This review summarizes a number of free-energy methods and how the latter were applied recently to the estimation of association free energies of CD inclusion complexes as a quantitative route towards an improved understanding of complexation processes.}, + timestamp = {2016-12-08T22:20:17Z}, + number = {6}, + journal = {Current Organic Chemistry}, + author = {Sheng Cai, Wen and Wang, Teng and Zhe Liu, Ying and Liu, Peng and Chipot, Christophe and Guang Shao, Xue}, + month = mar, + year = {2011}, + keywords = {complexation,cyclodextrins,free-energy calculations,Host-Guest Chemistry,molecular dynamics simulations}, + pages = {839--847}, + file = {IngentaConnect Full Text PDF:/Users/dmobley/Library/Application Support/Zotero/Profiles/i2jd8b87.default/zotero/storage/BK8V22JQ/Sheng Cai et al. - 2011 - Free Energy Calculations for Cyclodextrin Inclusio.pdf:application/pdf} +} + +@article{Wickstrom:2013:J.Chem.TheoryComput., + title = {Large {{Scale Affinity Calculations}} of {{Cyclodextrin Host}}\textendash{}{{Guest Complexes}}: {{Understanding}} the {{Role}} of {{Reorganization}} in the {{Molecular Recognition Process}}}, + volume = {9}, + issn = {1549-9618}, + shorttitle = {Large {{Scale Affinity Calculations}} of {{Cyclodextrin Host}}\textendash{}{{Guest Complexes}}}, + doi = {10.1021/ct400003r}, + abstract = {Host\textendash{}guest inclusion complexes are useful models for understanding the structural and energetic aspects of molecular recognition. Due to their small size relative to much larger protein\textendash{}ligand complexes, converged results can be obtained rapidly for these systems thus offering the opportunity to more reliably study fundamental aspects of the thermodynamics of binding. In this work, we have performed a large scale binding affinity survey of 57 $\beta$-cyclodextrin (CD) host\textendash{}guest systems using the binding energy distribution analysis method (BEDAM) with implicit solvation (OPLS-AA/AGBNP2). Converged estimates of the standard binding free energies are obtained for these systems by employing techniques such as parallel Hamiltonian replica exchange molecular dynamics, conformational reservoirs, and multistate free energy estimators. Good agreement with experimental measurements is obtained in terms of both numerical accuracy and affinity rankings. Overall, average effective binding energies reproduce affinity rank ordering better than the calculated binding affinities, even though calculated binding free energies, which account for effects such as conformational strain and entropy loss upon binding, provide lower root-mean-square errors when compared to measurements. Interestingly, we find that binding free energies are superior rank order predictors for a large subset containing the most flexible guests. The results indicate that, while challenging, accurate modeling of reorganization effects can lead to ligand design models of superior predictive power for rank ordering relative to models based only on ligand\textendash{}receptor interaction energies.}, + timestamp = {2016-12-08T22:20:35Z}, + number = {7}, + urldate = {2016-12-08}, + journal = {J. Chem. Theory Comput.}, + author = {Wickstrom, Lauren and He, Peng and Gallicchio, Emilio and Levy, Ronald M.}, + month = jul, + year = {2013}, + pages = {3136--3150}, + file = {ACS Full Text PDF w/ Links:/Users/dmobley/Library/Application Support/Zotero/Profiles/i2jd8b87.default/zotero/storage/HV9CMMQR/Wickstrom et al. - 2013 - Large Scale Affinity Calculations of Cyclodextrin .pdf:application/pdf;ACS Full Text Snapshot:/Users/dmobley/Library/Application Support/Zotero/Profiles/i2jd8b87.default/zotero/storage/7W3466H6/ct400003r.html:text/html} +} + +@article{Shi:2014:TheorChemAcc, + title = {Stereoselective Inclusion Mechanism of Ketoprofen into $\beta$-Cyclodextrin: Insights from Molecular Dynamics Simulations and Free Energy Calculations}, + volume = {133}, + issn = {1432-881X, 1432-2234}, + shorttitle = {Stereoselective Inclusion Mechanism of Ketoprofen into $\beta$-Cyclodextrin}, + doi = {10.1007/s00214-014-1556-8}, + abstract = {The host\textendash{}guest inclusion mechanism formed between $\beta$-cyclodextrin and those poorly water-soluble drug molecules has important applications in supramolecular chemistry, biology and pharmacy. In this work, the chiral recognition ability of $\beta$-cyclodextrin to one of nonsteroidal anti-inflammatory drugs, ketoprofen, has been systematically investigated using molecular dynamics and free energy simulation methods. The R- and S-enantiomers of ketoprofen were explicitly bound within the cyclodextrin cavity in our simulations, respectively. In consistent with experimental observations, tiny structural difference between two isomers could be observed. Calculated absolute binding free energies using adapted biasing force (ABF) method and MM/GBSA approach for both isomers are comparable to experimental values. Significant binding fluctuations along the MD trajectory have been observed. The free energy profiles calculated using two different approaches reveal that the ketoprofen prefers binding in the cavity with the carboxylate group facing the wider edge of $\beta$-cyclodextrin. Similar free energy profiles for two enantiomers obtained using ABF calculations indicate that it is very hard to separate and identify the chiral conjugates within the framework of the natural $\beta$-cyclodextrin.}, + language = {en}, + timestamp = {2016-12-08T22:20:47Z}, + number = {10}, + urldate = {2016-12-08}, + journal = {Theor Chem Acc}, + author = {Shi, Mingsong and Zhang, Chunchun and Xie, Yani and Xu, Dingguo}, + month = oct, + year = {2014}, + pages = {1556}, + file = {Full Text PDF:/Users/dmobley/Library/Application Support/Zotero/Profiles/i2jd8b87.default/zotero/storage/5HDB8JG6/Shi et al. - 2014 - Stereoselective inclusion mechanism of ketoprofen .pdf:application/pdf;Snapshot:/Users/dmobley/Library/Application Support/Zotero/Profiles/i2jd8b87.default/zotero/storage/DVS5C3PB/s00214-014-1556-8.html:text/html} +} + +@article{Zhang:2015:J.Chem.TheoryComput., + title = {Generalized {{Born}} and {{Explicit Solvent Models}} for {{Free Energy Calculations}} in {{Organic Solvents}}: {{Cyclodextrin Dimerization}}}, + volume = {11}, + issn = {1549-9618}, + shorttitle = {Generalized {{Born}} and {{Explicit Solvent Models}} for {{Free Energy Calculations}} in {{Organic Solvents}}}, + doi = {10.1021/acs.jctc.5b00620}, + abstract = {Evaluation of solvation (binding) free energies with implicit solvent models in different dielectric environments for biological simulations as well as high throughput ligand screening remain challenging endeavors. In order to address how well implicit solvent models approximate explicit ones we examined four generalized Born models (GBStill, GBHCT, GBOBCI, and GBOBCII) for determining the dimerization free energy ($\Delta$G0) of $\beta$-cyclodextrin monomers in 17 implicit solvents with dielectric constants (D) ranging from 5 to 80 and compared the results to previous free energy calculations with explicit solvents (Zhang et al. J. Phys. Chem. B 2012, 116, 12684-12693). The comparison indicates that neglecting the environmental dependence of Born radii appears acceptable for such calculations involving cyclodextrin and that the GBStill and GBOBCI models yield a reasonable estimation of $\Delta$G0, although the details of binding are quite different from explicit solvents. Large discrepancies between implicit and explicit solvent models occur in high-dielectric media with strong hydrogen bond (HB) interruption properties. $\Delta$G0 with the GB models is shown to correlate strongly to 2(D\textendash{}1)/(2D+1) (R2 $\sim$ 0.90) in line with the Onsager reaction field (Onsager J. Am. Chem. Soc. 1936, 58, 1486-1493) but to be very sensitive to D (D $<$ 10) as well. Both high-dielectric environments where hydrogen bonds are of interest and low-dielectric media such as protein binding pockets and membrane interiors therefore need to be considered with caution in GB-based calculations. Finally, a literature analysis of Gibbs energy of solvation of small molecules in organic liquids shows that the Onsager relation does not hold for real molecules since the correlation between $\Delta$G0 and 2(D\textendash{}1)/(2D+1) is low for most solutes. Interestingly, explicit solvent calculations of the solvation free energy (Zhang et al. J. Chem. Inf. Model. 2015, 55, 1192-1201) reproduce the weak experimental correlations with 2(D\textendash{}1)/(2D+1) very well.}, + timestamp = {2016-12-08T22:20:57Z}, + number = {11}, + urldate = {2016-12-08}, + journal = {J. Chem. Theory Comput.}, + author = {Zhang, Haiyang and Tan, Tianwei and {van der Spoel}, David}, + month = nov, + year = {2015}, + pages = {5103--5113}, + file = {ACS Full Text PDF w/ Links:/Users/dmobley/Library/Application Support/Zotero/Profiles/i2jd8b87.default/zotero/storage/AEA78Q3T/Zhang et al. - 2015 - Generalized Born and Explicit Solvent Models for F.pdf:application/pdf;ACS Full Text Snapshot:/Users/dmobley/Library/Application Support/Zotero/Profiles/i2jd8b87.default/zotero/storage/WBXAKM36/acs.jctc.html:text/html} +} + +@article{Khuntawee:2016:CarbohydratePolymers, + title = {Conformation Study of Ɛ-Cyclodextrin: {{Replica}} Exchange Molecular Dynamics Simulations}, + volume = {141}, + issn = {0144-8617}, + shorttitle = {Conformation Study of Ɛ-Cyclodextrin}, + doi = {10.1016/j.carbpol.2015.10.018}, + abstract = {There is growing interest in large-ring cyclodextrins (LR-CDs) which are known to be good host molecules for larger ligands. The isolation of a defined size LR-CD is an essential prerequisite for studying their structural properties. Unfortunately the purification procedure of these substances turned out to be very laborious. Finally the problem could be circumvented by a theoretical consideration: the highly advantageous replica exchange molecular dynamics (REMD) simulation (particularly suitable for studies of conformational changes) offers an ideal approach for studying the conformational change of $\varepsilon$-cyclodextrin (CD10), a smaller representative of LR-CDs. Three carbohydrate force fields and three solvent models were tested. The conformational behavior of CD10 was analyzed in terms of the flip (turn) of the glucose subunits within the macrocyclic ring. In addition a ranking of conformations with various numbers of turns was preformed. Our findings might be also helpful in the temperature controlled synthesis of LR-CDs as well as other experimental conditions, in particular for the host\textendash{}guest reaction.}, + timestamp = {2016-12-08T22:22:27Z}, + urldate = {2016-12-08}, + journal = {Carbohydrate Polymers}, + author = {Khuntawee, Wasinee and Rungrotmongkol, Thanyada and Wolschann, Peter and Pongsawasdi, Piamsook and Kungwan, Nawee and Okumura, Hisashi and Hannongbua, Supot}, + month = may, + year = {2016}, + keywords = {Conformational changes,ɛ-Cyclodextrin,Large-ring cyclodextrin,Replica exchange molecular dynamics}, + pages = {99--105}, + file = {ScienceDirect Full Text PDF:/Users/dmobley/Library/Application Support/Zotero/Profiles/i2jd8b87.default/zotero/storage/DH4662HX/Khuntawee et al. - 2016 - Conformation study of ɛ-cyclodextrin Replica exch.pdf:application/pdf;ScienceDirect Snapshot:/Users/dmobley/Library/Application Support/Zotero/Profiles/i2jd8b87.default/zotero/storage/ASMB8GNQ/S0144861715009984.html:text/html} +} + +@article{Gebhardt:2016:FluidPhaseEquilibria, + series = {Special issue on Biothermodynamics}, + title = {Calculation of Binding Affinities for Linear Alcohols to $\alpha$-Cyclodextrin by Twin-System Enveloping Distribution Sampling Simulations}, + volume = {422}, + issn = {0378-3812}, + doi = {10.1016/j.fluid.2016.02.001}, + abstract = {Comparison of three variants of the GROMOS force field for carbohydrates, 53A6GLYC, 56A6CARBO\_R and a small modification of the latter, referred to as 56A6CARBO\_R+, has been conducted by examining the standard binding free enthalpy of linear alcohols from 1-butanol to 1-dodecanol to $\alpha$-cyclodextrin in aqueous solution. The double decoupling approach was employed to calculate the standard binding free enthalpy of 1-hexanol, while for all other compounds relative binding free enthalpies for pairs of alcohols differing by two carbon atoms were calculated from twin-system enveloping distribution sampling simulations. While all three force fields slightly overestimate the magnitude of the standard binding free enthalpy of 1-hexanol the 53A6GLYC and 56A6CARBO\_R+ parameter sets are a little closer to the experimental data than 56A6CARBO\_R. For relative binding free enthalpies no significant difference between the force fields was observed. All three force fields support those experimental measurements that found a distinct increase in the binding affinity with increasing number of carbon atoms even for longer chain lengths up to 1-dodecanol.}, + timestamp = {2016-12-08T22:22:56Z}, + urldate = {2016-12-08}, + journal = {Fluid Phase Equilibria}, + author = {Gebhardt, Julia and Hansen, Niels}, + month = aug, + year = {2016}, + keywords = {Cyclodextrin,force field,Free energy,Molecular dynamics}, + pages = {1--17}, + file = {ScienceDirect Full Text PDF:/Users/dmobley/Library/Application Support/Zotero/Profiles/i2jd8b87.default/zotero/storage/UVA9NBCA/Gebhardt and Hansen - 2016 - Calculation of binding affinities for linear alcoh.pdf:application/pdf;ScienceDirect Snapshot:/Users/dmobley/Library/Application Support/Zotero/Profiles/i2jd8b87.default/zotero/storage/2EV4F4WG/S0378381216300516.html:text/html} +} + +@article{Kirschner:2008:J.Comput.Chem., + title = {{{GLYCAM06}}: {{A}} Generalizable Biomolecular Force Field. {{Carbohydrates}}}, + volume = {29}, + issn = {1096-987X}, + shorttitle = {{{GLYCAM06}}}, + doi = {10.1002/jcc.20820}, + abstract = {A new derivation of the GLYCAM06 force field, which removes its previous specificity for carbohydrates, and its dependency on the AMBER force field and parameters, is presented. All pertinent force field terms have been explicitly specified and so no default or generic parameters are employed. The new GLYCAM is no longer limited to any particular class of biomolecules, but is extendible to all molecular classes in the spirit of a small-molecule force field. The torsion terms in the present work were all derived from quantum mechanical data from a collection of minimal molecular fragments and related small molecules. For carbohydrates, there is now a single parameter set applicable to both $\alpha$- and $\beta$-anomers and to all monosaccharide ring sizes and conformations. We demonstrate that deriving dihedral parameters by fitting to QM data for internal rotational energy curves for representative small molecules generally leads to correct rotamer populations in molecular dynamics simulations, and that this approach removes the need for phase corrections in the dihedral terms. However, we note that there are cases where this approach is inadequate. Reported here are the basic components of the new force field as well as an illustration of its extension to carbohydrates. In addition to reproducing the gas-phase properties of an array of small test molecules, condensed-phase simulations employing GLYCAM06 are shown to reproduce rotamer populations for key small molecules and representative biopolymer building blocks in explicit water, as well as crystalline lattice properties, such as unit cell dimensions, and vibrational frequencies. \textcopyright{} 2007 Wiley Periodicals, Inc. J Comput Chem, 2008}, + language = {en}, + timestamp = {2016-12-08T22:23:05Z}, + number = {4}, + urldate = {2016-12-08}, + journal = {J. Comput. Chem.}, + author = {Kirschner, Karl N. and Yongye, Austin B. and Tschampel, Sarah M. and Gonz{\'a}lez-Outeiri{\~n}o, Jorge and Daniels, Charlisa R. and Foley, B. Lachele and Woods, Robert J.}, + month = mar, + year = {2008}, + keywords = {Amber,carbohydrate,force field,GLYCAM,Molecular dynamics,parameter development}, + pages = {622--655}, + file = {Full Text PDF:/Users/dmobley/Library/Application Support/Zotero/Profiles/i2jd8b87.default/zotero/storage/W4VHGZZV/Kirschner et al. - 2008 - GLYCAM06 A generalizable biomolecular force field.pdf:application/pdf;Snapshot:/Users/dmobley/Library/Application Support/Zotero/Profiles/i2jd8b87.default/zotero/storage/TZ35RZUU/abstract.html:text/html} +} + +@article{Cezard:2011:PhysicalChemistryChemicalPhysics, + title = {Molecular Dynamics Studies of Native and Substituted Cyclodextrins in Different Media: 1. {{Charge}} Derivation and Force Field Performances}, + volume = {13}, + shorttitle = {Molecular Dynamics Studies of Native and Substituted Cyclodextrins in Different Media}, + doi = {10.1039/C1CP20854C}, + language = {en}, + timestamp = {2016-12-08T22:24:06Z}, + number = {33}, + urldate = {2016-12-08}, + journal = {Physical Chemistry Chemical Physics}, + author = {C{\'e}zard, Christine and Trivelli, Xavier and Aubry, Fr{\'e}d{\'e}ric and Djeda{\"\i}ni-Pilard, Florence and Dupradeau, Fran{\c c}ois-Yves}, + year = {2011}, + pages = {15103--15121}, + file = {Full Text PDF:/Users/dmobley/Library/Application Support/Zotero/Profiles/i2jd8b87.default/zotero/storage/KZQ9B7HH/Cézard et al. - 2011 - Molecular dynamics studies of native and substitut.pdf:application/pdf;Snapshot:/Users/dmobley/Library/Application Support/Zotero/Profiles/i2jd8b87.default/zotero/storage/B579VHJU/c1cp20854c.html:text/html} +} + +@article{Guvench:2011:J.Chem.TheoryComput., + title = {{{CHARMM Additive All}}-{{Atom Force Field}} for {{Carbohydrate Derivatives}} and {{Its Utility}} in {{Polysaccharide}} and {{Carbohydrate}}\textendash{}{{Protein Modeling}}}, + volume = {7}, + issn = {1549-9618}, + doi = {10.1021/ct200328p}, + abstract = {Monosaccharide derivatives such as xylose, fucose, N-acetylglucosamine (GlcNAc), N-acetylgalactosamine (GlaNAc), glucuronic acid, iduronic acid, and N-acetylneuraminic acid (Neu5Ac) are important components of eukaryotic glycans. The present work details the development of force-field parameters for these monosaccharides and their covalent connections to proteins via O linkages to serine or threonine side chains and via N linkages to asparagine side chains. The force field development protocol was designed to explicitly yield parameters that are compatible with the existing CHARMM additive force field for proteins, nucleic acids, lipids, carbohydrates, and small molecules. Therefore, when combined with previously developed parameters for pyranose and furanose monosaccharides, for glycosidic linkages between monosaccharides, and for proteins, the present set of parameters enables the molecular simulation of a wide variety of biologically important molecules such as complex carbohydrates and glycoproteins. Parametrization included fitting to quantum mechanical (QM) geometries and conformational energies of model compounds, as well as to QM pair interaction energies and distances of model compounds with water. Parameters were validated in the context of crystals of relevant monosaccharides, as well NMR and/or X-ray crystallographic data on larger systems including oligomeric hyaluronan, sialyl Lewis X, O- and N-linked glycopeptides, and a lectin:sucrose complex. As the validated parameters are an extension of the CHARMM all-atom additive biomolecular force field, they further broaden the types of heterogeneous systems accessible with a consistently developed force-field model.}, + timestamp = {2016-12-08T22:24:15Z}, + number = {10}, + urldate = {2016-12-08}, + journal = {J. Chem. Theory Comput.}, + author = {Guvench, Olgun and Mallajosyula, Sairam S. and Raman, E. Prabhu and Hatcher, Elizabeth and Vanommeslaeghe, Kenno and Foster, Theresa J. and Jamison, Francis W. and MacKerell, Alexander D.}, + month = oct, + year = {2011}, + pages = {3162--3180}, + file = {ACS Full Text PDF w/ Links:/Users/dmobley/Library/Application Support/Zotero/Profiles/i2jd8b87.default/zotero/storage/NP5X3XZX/Guvench et al. - 2011 - CHARMM Additive All-Atom Force Field for Carbohydr.pdf:application/pdf;ACS Full Text Snapshot:/Users/dmobley/Library/Application Support/Zotero/Profiles/i2jd8b87.default/zotero/storage/B45MM42W/ct200328p.html:text/html} +} + +@article{Xiong:2015:CarbohydrateResearch, + title = {Force Fields and Scoring Functions for Carbohydrate Simulation}, + volume = {401}, + issn = {0008-6215}, + doi = {10.1016/j.carres.2014.10.028}, + abstract = {Carbohydrate dynamics plays a vital role in many biological processes, but we are not currently able to probe this with experimental approaches. The highly flexible nature of carbohydrate structures differs in many aspects from other biomolecules, posing significant challenges for studies employing computational simulation. Over past decades, computational study of carbohydrates has been focused on the development of structure prediction methods, force field optimization, molecular dynamics simulation, and scoring functions for carbohydrate\textendash{}protein interactions. Advances in carbohydrate force fields and scoring functions can be largely attributed to enhanced computational algorithms, application of quantum mechanics, and the increasing number of experimental structures determined by X-ray and NMR techniques. The conformational analysis of carbohydrates is challengeable and has gone into intensive study in elucidating the anomeric, the exo-anomeric, and the gauche effects. Here, we review the issues associated with carbohydrate force fields and scoring functions, which will have a broad application in the field of carbohydrate-based drug design.}, + timestamp = {2016-12-08T22:24:41Z}, + urldate = {2016-12-08}, + journal = {Carbohydrate Research}, + author = {Xiong, Xiuming and Chen, Zhaoqiang and Cossins, Benjamin P. and Xu, Zhijian and Shao, Qiang and Ding, Kai and Zhu, Weiliang and Shi, Jiye}, + month = jan, + year = {2015}, + keywords = {Carbohydrate force field,Carbohydrate–protein interaction,drug design,Scoring function,simulation}, + pages = {73--81}, + file = {ScienceDirect Full Text PDF:/Users/dmobley/Library/Application Support/Zotero/Profiles/i2jd8b87.default/zotero/storage/3AFXIBBG/Xiong et al. - 2015 - Force fields and scoring functions for carbohydrat.pdf:application/pdf;ScienceDirect Snapshot:/Users/dmobley/Library/Application Support/Zotero/Profiles/i2jd8b87.default/zotero/storage/CES6PKWF/S0008621514004017.html:text/html} +} + +@inproceedings{Mic:2013:AIPConferenceProceedings, + title = {Inclusion Complex of Benzocaine and $\beta$-Cyclodextrin: {{1H NMR}} and Isothermal Titration Calorimetry Studies}, + volume = {1565}, + shorttitle = {Inclusion Complex of Benzocaine and $\beta$-Cyclodextrin}, + doi = {10.1063/1.4833697}, + abstract = {The supramolecular structure of the inclusion complex of $\beta$-cyclodextrin with benzocaine in aqueous solution has been investigated by 1 H NMR spectroscopy and isothermal titration nanocalorimetry (ITC). Analysis of 1 H NMR data by continuous variation method indicates that the benzocaine: $\beta$-cyclodextrin inclusion complex occurs and has a 1:1 stoichiometry. Rotating frame NOE spectroscopy (ROESY) was used to ascertain the solution geometry of the host-guest complex which indicates that the benzocaine molecule was included with the aromatic ring into the cyclodextrin cavity. Although the affinity of benzocaine for cyclodextrin is relatively high, the association constant cannot be measured using ITC due to the low solubility of benzocaine in water.}, + timestamp = {2016-12-08T22:25:28Z}, + urldate = {2016-12-08}, + booktitle = {{{AIP Conference Proceedings}}}, + publisher = {{AIP Publishing}}, + author = {Mic, Mihaela and P\i\^rn{\u a}u, Adrian and Bogdan, Mircea and Turcu, Ioan}, + month = nov, + year = {2013}, + keywords = {Calorimeters,Data analysis,Differential scanning calorimeters,Equilibrium constants,Nuclear magnetic resonance}, + pages = {63--66}, + file = {Full Text PDF:/Users/dmobley/Library/Application Support/Zotero/Profiles/i2jd8b87.default/zotero/storage/7PWVH9XH/Mic et al. - 2013 - Inclusion complex of benzocaine and β-cyclodextrin.pdf:application/pdf;Snapshot:/Users/dmobley/Library/Application Support/Zotero/Profiles/i2jd8b87.default/zotero/storage/AKBMK5J2/1.html:text/html} +} + +@book{Dodziuk:2006:, + title = {Cyclodextrins and {{Their Complexes}}: {{Chemistry}}, {{Analytical Methods}}, {{Applications}}}, + timestamp = {2016-12-08T22:26:43Z}, + publisher = {{John Wiley \& Sons}}, + author = {Dodziuk, H}, + year = {2006} +} + +@incollection{Fromming:1994:CyclodextrinsinPharmacy, + title = {Cyclodextrin {{Derivatives}}}, + timestamp = {2016-12-08T22:29:46Z}, + booktitle = {Cyclodextrins in {{Pharmacy}}}, + author = {Fr{\"o}mming, K.-H. and Szejtli, J{\'o}zsef}, + year = {1994}, + pages = {19--32} +} + +@article{Gilson:2016:Nucl.AcidsRes., + title = {{{BindingDB}} in 2015: {{A}} Public Database for Medicinal Chemistry, Computational Chemistry and Systems Pharmacology}, + volume = {44}, + issn = {0305-1048, 1362-4962}, + shorttitle = {{{BindingDB}} in 2015}, + doi = {10.1093/nar/gkv1072}, + abstract = {BindingDB, www.bindingdb.org, is a publicly accessible database of experimental protein-small molecule interaction data. Its collection of over a million data entries derives primarily from scientific articles and, increasingly, US patents. BindingDB provides many ways to browse and search for data of interest, including an advanced search tool, which can cross searches of multiple query types, including text, chemical structure, protein sequence and numerical affinities. The PDB and PubMed provide links to data in BindingDB, and vice versa; and BindingDB provides links to pathway information, the ZINC catalog of available compounds, and other resources. The BindingDB website offers specialized tools that take advantage of its large data collection, including ones to generate hypotheses for the protein targets bound by a bioactive compound, and for the compounds bound by a new protein of known sequence; and virtual compound screening by maximal chemical similarity, binary kernel discrimination, and support vector machine methods. Specialized data sets are also available, such as binding data for hundreds of congeneric series of ligands, drawn from BindingDB and organized for use in validating drug design methods. BindingDB offers several forms of programmatic access, and comes with extensive background material and documentation. Here, we provide the first update of BindingDB since 2007, focusing on new and unique features and highlighting directions of importance to the field as a whole.}, + language = {en}, + timestamp = {2016-12-08T22:45:13Z}, + number = {D1}, + urldate = {2016-12-08}, + journal = {Nucl. Acids Res.}, + author = {Gilson, Michael K. and Liu, Tiqing and Baitaluk, Michael and Nicola, George and Hwang, Linda and Chong, Jenny}, + month = apr, + year = {2016}, + pages = {D1045--D1053}, + file = {Full Text PDF:/Users/dmobley/Library/Application Support/Zotero/Profiles/i2jd8b87.default/zotero/storage/X65QJVGV/Gilson et al. - 2016 - BindingDB in 2015 A public database for medicinal.pdf:application/pdf;Snapshot:/Users/dmobley/Library/Application Support/Zotero/Profiles/i2jd8b87.default/zotero/storage/QSEKC6IN/D1045.html:text/html}, + pmid = {26481362} +} + @comment{jabref-meta: groupsversion:3;} @comment{jabref-meta: groupstree: 0 AllEntriesGroup:; -1 ExplicitGroup:bromodomains\;0\;aldeghi_accurate_2016\;; +1 ExplicitGroup:bromodomains\;0\;aldeghi_accurate_2016\;Filippakopoulo +s:2010:Nature\;Chung:2011:J.Med.Chem.\;Hewings:2012:J.Med.Chem.\;Bambo +rough:2015:J.Med.Chem.\;Ran:2015:J.Med.Chem.\;Zhang:2013:J.Med.Chem.\; +Gosmini:2014:J.Med.Chem.\;Yang:2016:BioorganicChemistry\;; 1 ExplicitGroup:CB7\;0\;henriksen_computational_2015\;monroe_convergin g_2014\;fenley_bridging_2014\;yin_toward_2015\;muddana_blind_2014\;mog haddam_new_2011\;gilson_stress_2010\;moghaddam_hostguest_2009\;wyman_c @@ -3069,12 +3913,27 @@ @comment{jabref-meta: ao_attomolar_2014\;Jang:2014:Angew.Chem.Int.Ed.\;freeman_cucurbituril_ 1981\;velez-vega_force_2012\;velez-vega_overcoming_2013\;Cong:2016:Org .Biomol.Chem.\;vinciguerra_synthesis_2015\;assaf_cucurbiturils:_2015\; -Lee:2003:Acc.Chem.Res.\;; +Lee:2003:Acc.Chem.Res.\;Marquez:2004:J.Am.Chem.Soc.\;; 1 ExplicitGroup:CCP\;0\;brenk_probing_2006\;rocklin_blind_2013\;banba_ efficient_2000\;banba_free_2000\;rosenfeld_excision_2002\;musah_artifi -cial_2002\;fitzgerald_ligand-gated_1996\;; +cial_2002\;fitzgerald_ligand-gated_1996\;Fitzgerald:1995:ProteinScienc +e\;; 1 ExplicitGroup:cyclodextrin\;0\;henriksen_computational_2015\;wickstr -om_parameterization_2016\;; +om_parameterization_2016\;rekharsky_complexation_1998\;Connors:1997:Ch +em.Rev.\;Carrazana:2005:J.Phys.Chem.B\;Cotner:1998:J.Org.Chem.\;Wszela +ka-Rylik:2013:JThermAnalCalorim\;Shu:2007:BritishJournalofPharmacology +\;Rodriguez-Perez:2006:J.Pharm.Sci.\;Szente:1999:AdvancedDrugDeliveryR +eviews\;Qu:2002:JournalofInclusionPhenomena\;Jindrich:2005:J.Org.Chem. +\;Faugeras:2012:Eur.J.Org.Chem.\;Mark:1994:J.Am.Chem.Soc.\;Luzhkov:199 +9:ChemicalPhysicsLetters\;Bea:2002:TheorChemAcc\;Chen:2004:Biophysical +Journal\;Sellner:2008:J.Phys.Chem.B\;Cai:2009:J.Phys.Chem.B\;ShengCai: +2011:CurrentOrganicChemistry\;Wickstrom:2013:J.Chem.TheoryComput.\;Shi +:2014:TheorChemAcc\;Zhang:2015:J.Chem.TheoryComput.\;Khuntawee:2016:Ca +rbohydratePolymers\;Gebhardt:2016:FluidPhaseEquilibria\;Kirschner:2008 +:J.Comput.Chem.\;Cezard:2011:PhysicalChemistryChemicalPhysics\;Guvench +:2011:J.Chem.TheoryComput.\;Xiong:2015:CarbohydrateResearch\;Mic:2013: +AIPConferenceProceedings\;Dodziuk:2006:\;Fromming:1994:Cyclodextrinsin +Pharmacy\;; 1 ExplicitGroup:FKBP\;0\;jayachandran_parallelized-over-parts_2006\;yt reberg_absolute_2009\;fujitani_direct_2005\;fujitani_massively_2009\;w ang_absolute_2006\;lee_calculation_2006\;shirts_calculating_2004\;; @@ -3108,7 +3967,9 @@ @comment{jabref-meta: ow_2015\;; 1 ExplicitGroup:thrombin\;0\;wang_achieving_2012\;schrodinger_accurate _2015\;calabro_elucidation_2016\;baum_non-additivity_2010\;calabro_acc -elerating_2015\;; +elerating_2015\;StefanicAnderluh:2005:J.Med.Chem.\;Ueno:2005:Bioorgani +c&MedicinalChemistryLetters\;Putta:2005:J.Med.Chem.\;Dullweber:2001:Jo +urnalofMolecularBiology\;; 1 ExplicitGroup:trypsin\;0\;plattner_protein_2015\;talhout_understandi ng_2003\;skillman_sampl3_2012\;Newman:2011:JComputAidedMolDes\;de_ruit er_efficient_2012\;jiao_calculation_2008\;villa_sampling_2003\;jiao_tr diff --git a/paper/benchmarkset.pdf b/paper/benchmarkset.pdf index 235f074..1f883dc 100644 Binary files a/paper/benchmarkset.pdf and b/paper/benchmarkset.pdf differ diff --git a/paper/benchmarkset.tex b/paper/benchmarkset.tex index d853385..43db580 100644 --- a/paper/benchmarkset.tex +++ b/paper/benchmarkset.tex @@ -48,7 +48,6 @@ \title{Predicting binding free energies: Frontiers and benchmarks} - %Authors, affiliations, address. \author{David L. Mobley} \email{dmobley@mobleylab.org} @@ -58,12 +57,27 @@ \affiliation{Skaggs School of Pharmacy and Pharmaceutical Sciences, University of California, San Diego, CA, USA, 92092} +%%% DATE: +% For ongoing development use +%\date{\today} +% For release versions use +\date{ December 8, 2016 } + \begin{abstract} Binding free energy calculations based on molecular simulations provide predicted affinities for biomolecular complexes. These calculations begin with a detailed description of a system, including its chemical composition and the interactions between its components. -Simulations of the system are then used to compute thermodynamic information, such as binding affinities. Because of their promise for guiding molecular design, these calculations have recently begun to see widespread applications in early stage drug discovery. However, many challenges remain to make them a robust and reliable tool. Here, we briefly explain how the calculations work, highlight key challenges, and argue for the development of accepted benchmark test systems that will help the research community generate and evaluate progress. +Simulations of the system are then used to compute thermodynamic information, such as binding affinities. +Because of their growing promise for guiding molecular design, these calculations have recently begun to see widespread applications in early stage drug discovery. However, many challenges remain to make them a robust and reliable tool. +Here, we provide an overview of these methods, discuss determinants of accuracy and precision, highlight key challenges, and argue for the development of accepted benchmark test systems that will help the research community generate and evaluate progress. +\\ +\\ +%PUT MANUSCRIPT VERSION HERE +% Use this style for ongoing development: "Version 1.0.5 pre-release" +% Use this style for releases: "Version 2.0" +{\bf Manuscript version 1.1.} See \url{https://github.com/mobleylab/benchmarksets} for all versions. \end{abstract} + \maketitle @@ -90,7 +104,7 @@ \subsection{Imagining a tool for drug discovery} \subsection{Increasing accuracy will yield increasing payoffs} -Recent progress in computational power, especially the widespread availability of graphics processing units (GPUs) and advances in automation~\cite{liu_lead_2013} and sampling protocols, have helped simulation-based techniques reach the point where they now appear to have sufficient accuracy to be genuinely useful in guiding pharmaceutical drug discovery at least for a certain subset of problems~\cite{mikulskis_large-scale_2014, homeyer_binding_2014, sherborne_preprint_2016, schrodinger_accurate_2015, christ_binding_2016, cui_affinity_2016, verras_free_2016}. +Recent progress in computational power, especially the widespread availability of graphics processing units (GPUs) and advances in automation~\cite{liu_lead_2013} and sampling protocols, have helped simulation-based techniques reach the point where they now appear to have sufficient accuracy to be genuinely useful in guiding pharmaceutical drug discovery, at least for a certain subset of problems~\cite{mikulskis_large-scale_2014, homeyer_binding_2014, sherborne_preprint_2016, schrodinger_accurate_2015, christ_binding_2016, cui_affinity_2016, verras_free_2016}. Specifically, in some situations, free energy calculations appear to be capable of achieving RMS errors in the 1-2 kcal/mol range with current force fields, even in prospective applications. As a consequence, pharmaceutical companies are beginning to use these methods in discovery projects. The most immediate application of these techniques is to guide synthesis for lead optimization, but applications to scaffold hopping and in other areas also appear possible. @@ -102,7 +116,7 @@ \subsection{Increasing accuracy will yield increasing payoffs} Thus, one can gain substantial benefit from simulations that are good yet still quite imperfect. % Following paragraph could be cut for space reasons if needed. -More broadly, this analysis does not address the net value of computational affinity predictions in drug discovery. +More broadly, though, this analysis does not address the net value of computational affinity predictions in drug discovery. Costs include those of the software, computer time, and personnel required to incorporate calculations into the workflow; while benefits include the savings, revenue gains, and externalities attributable to reducing the number of low-affinity compounds synthesized and arriving earlier at a potent drug candidate. In addition, with sufficiently reliable predictions, chemists may choose to tackle difficult synthesis efforts they otherwise might have avoided, resulting in more novel and valuable chemical matter. @@ -113,14 +127,14 @@ \subsection{Overview of free energy calculations} In either case, the energy of a given configuration is provided by a potential function, or force field, which estimates the potential energy of a system of solute and solvent molecules as a function of the coordinates of all of its atoms. Such simulations may be used in several different ways to compute binding free energies or relative binding free energies, as detailed elsewhere~\cite{michel_prediction_2010, christ_basic_2010, chodera_alchemical_2011, shirts_introduction_2013} and summarized below. In all cases, however, the calculations yield the free energy difference between two states of a molecular system, and they do so by computing the reversible work for changing the initial state to the final one. Two broad approaches deserve mention. -The first general approach directly computes the standard free energy of binding of two molecules by computing the reversible work of transferring the ligand from the binding site into solution. +The first general approach directly computes the standard free energy of binding of two molecules via evaluating the reversible work of transferring the ligand from the binding site into solution. (This is sometimes called an absolute binding free energy calculation.) The pathway of this change may be one that is physically realizable, or one that is only realizable {\em in silico}, in which case it is sometimes called an ``alchemical'' pathway. Physical pathway methods provide the standard binding free energy by computing the reversible work of, in effect, pulling the ligand out of the binding site. Although, by definition, the pathway used must be a physical one that could occur in nature, it need not be probable, and improbable pathways, governed by an order parameter specifying how far the ligand is from the binding site, are often used~\cite{woo_calculation_2005, ytreberg_absolute_2009, velez-vega_overcoming_2013, henriksen_computational_2015, hsiao_prediction_2014, bhakat_resolving_2016}. In addition, artificial restraints may be useful to avoid sampling problems in the face of often complex barriers along the pathway ~\cite{woo_calculation_2005, velez-vega_overcoming_2013, henriksen_computational_2015, hsiao_prediction_2014, bhakat_resolving_2016}. By contrast, alchemical pathway methods artificially decouple the ligand from the binding site and then recouple it to solution from the protein~\cite{jorgensen_efficient_1988, Hermans:1986:Isr.J.Chem., gilson_statistical-thermodynamic_1997, boresch_absolute_2003, mobley_use_2006}. -Although alchemical decoupling methods may avoid clashes of the ligand with the protein that might be problematic in pathway methods for a tight binding site, they still can pose some of the same sampling challenges. +Although alchemical decoupling methods may avoid clashes of the ligand with the protein that might be problematic when pathway methods are applied to a protein with a buried binding site, they still can pose some of the same sampling challenges. For example, sampling of the unbound receptor must be adequate after the ligand is removed, and water molecules must have time to equilibrate in the vacated binding site. Given that free energy is a state function, it is not surprising that alchemical and physical pathway approaches yield apparently comparable results when applied to the same systems~\cite{lee_calculation_2006, gumbart_standard_2013, de_ruiter_proteinligand_2013, yin_overview_2016}. @@ -129,11 +143,11 @@ \subsection{Overview of free energy calculations} In addition, there may be concerns about slow conformational relaxation of the protein in response to the change in ligand. Nonetheless, alchemical relative free energy calculations currently are the best automated and most widely used free energy methods~\cite{mobley_perspective_2012, liu_lead_2013, schrodinger_accurate_2015}. -Importantly, the accuracy and precision of all of these methods are controlled by the same considerations. +The accuracy and precision of all of these methods are controlled by three major considerations. First, many conformations typically need to be generated, or sampled, in order to obtain an adequate representation of the Boltzmann distribution. In the limit of infinite sampling, a correctly implemented method would yield the single value of the free energy difference dictated by the specification of the molecular system and the chosen force field. In reality, however, only finite sampling is possible, so the reported free energy will differ from the nominal value associated with infinite sampling. -In addition, because sampling methods are typically stochastic and the dynamics of molecular systems are highly sensitive to initial conditions~\cite{allen_computer_1989}, repeated calculations, using different random number seeds or initial states, will yield different results. +In addition, because sampling methods are typically stochastic and the dynamics of molecular systems are sensitive to initial conditions~\cite{allen_computer_1989}, repeated calculations, using different random number seeds or initial states, will yield somewhat different results. The problem of finite sampling is most acute for systems where low-energy (hence highly occupied) conformational states are separated by high effective barriers, whether energetic or entropic. Second, even if adequate sampling is achievable, free energy differences may disagree substantially with experiment if the force field is not sufficiently accurate. Third, errors may also arise if the representation of the system in the simulation does not adequately represent the actual system, e.g. if protonation states are assigned incorrectly and held fixed. @@ -145,14 +159,14 @@ \subsection{Challenges and the domain of applicability} \item a high quality receptor structure is available, without missing loops or other major uncertainties \item the protonation state of the ligand and binding-site residues (as well as any other relevant residues) can reliably be inferred \item the ligand binding mode is defined by crystallographic studies and is not expected to change much on modification -\item the receptor does not undergo substantial or slow conformational changes +\item the receptor does not undergo substantial, slow conformational changes \item key interactions are expected to be well-described by underlying force fields \end{itemize} \vspace{2mm} Beyond this domain of applicability---whose dimensions are, in fact, still somewhat vague~\cite{Abel:2016:vertex} --- substantial challenges may be encountered. For example, binding free energy calculations for a cytochrome C peroxidase mutant suggest limitations of fixed-charge force fields. In this case, the strength of electrostatic interactions in a buried, relatively nonpolar binding site appears to be overestimated by a conventional fixed-charge force field, likely due to underestimation of polarization effects~\cite{rocklin_blind_2013}. -Sampling problems are also common, with slow sidechain rearrangements and ligand binding mode rearrangements in model binding sites in T4 lysozyme posing timescale problems unless enhanced or biased sampling methods are carefully applied~\cite{mobley_confine_2007, boyce_predicting_2009, mobley_predicting_2007, jiang_free_2010, gallicchio_binding_2010, wang_achieving_2012}; and larger-scale protein motions induced by some ligands also posing challenges~\cite{boyce_predicting_2009, lim_sensitivity_2016}. +Sampling problems are also common, with slow sidechain rearrangements and ligand binding mode rearrangements in model binding sites in T4 lysozyme posing timescale problems unless enhanced or biased sampling methods are carefully applied~\cite{mobley_confine_2007, boyce_predicting_2009, mobley_predicting_2007, jiang_free_2010, gallicchio_binding_2010, wang_achieving_2012}; and larger-scale protein motions induced by some ligands can also pose challenges~\cite{boyce_predicting_2009, lim_sensitivity_2016}. Although such problems need not prevent free energy calculations from being used, they can require specific adjustment of procedures and parameters based on experience and knowledge of the system at hand. Thus, a key challenge for the field is how to use insights from well-studied cases to enable automation and reduce the detailed knowledge of each system required to carry out high quality simulations. @@ -184,7 +198,7 @@ \subsubsection{Hard benchmarks} \paragraph{Systems to test software implementations and usage} It is crucial yet nontrivial to validate that a simulation package correctly implements and applies the desired methods~\cite{shirts_lessons_2016}, and benchmark cases can help with this. -First, all software packages could be tested for their ability to generate correct potential energies for a single configuration of the specified molecular system and force field. +First, all software packages could be tested for their ability to generate correct potential energies for a single configuration of a specified molecular system and force field. These results should be correct to within rounding error and the precision of the physical constants used in the calculations~\cite{shirts_lessons_2016}. Similarly, different methods and software packages should give consistent binding free energies when identical force fields are applied with identical simulation setups and compositions. The benchmark systems for such testing can be simple and easy to converge, and high precision free energies (e.g., uncertainty $\approx 0.1$ kcal/mol) should serve as a reference. @@ -197,8 +211,8 @@ \subsubsection{Hard benchmarks} This sampling is typically done by running molecular dynamics simulations, and for systems as large and complex as proteins, it is difficult to carry out long enough simulations. Calculations with inadequate sampling yield results that are imprecise, in the sense that multiple independent calculations with slightly different initial conditions will yield significantly different results, and these ill-converged results will in general be poor estimates of the ideal result obtained in the limit of infinite sampling. Advanced simulation methods have been developed to speed convergence~\cite{tai_conformational_2004, shirts_introduction_2013}, but it is not always clear how various methods compare to one another. -To effectively compare such enhanced sampling methods, we need benchmark molecular systems, parameterized with a force field that many software packages can use, that embody various sampling challenges, such as high dimensionality and energetic and entropic barriers between highly occupied states, but which are just tractable enough that reliable results are available via suitable reference calculations. -Again, experimental data are not required, and the point of comparison may be, at least in part, sampling measures. +To effectively compare such enhanced sampling methods, we need benchmark molecular systems, parameterized with a force field that many software packages can use, that embody various sampling challenges, such as high dimensionality, and energetic and entropic barriers between highly occupied states, but which are just tractable enough that reliable results are available via suitable reference calculations. +Again, experimental data are not required, as the main purpose is to test and compare the power of various enhanced sampling technologies. \paragraph{Systems to assess force field accuracy} \label{pgph:accuracy} @@ -208,7 +222,7 @@ \subsubsection{Hard benchmarks} simple molecular recognition systems~\cite{henriksen_computational_2015}, as further discussed below. In such cases, absent complications like uncertain protonation states, the level of agreement with experiment reports directly on the accuracy of the force field. Thus, simple molecular recognition systems with reliable experimental binding data represent another valuable class of benchmarks. -Here, of course, experimental data are needed. Ideally, the physical materials will be fairly easy to obtain so that measurements can be replicated or new experimental conditions (such as temperature and solvent composition) explored. +Here, of course, experimental data are needed. Ideally, the physical materials will be fairly easy to obtain so that measurements can be replicated, and new experimental conditions, such as different temperatures and solvent compositions, can be explored. \subsubsection{Soft benchmarks} \paragraph{Systems to challenge conformational sampling techniques} @@ -219,15 +233,15 @@ \subsubsection{Soft benchmarks} Developers should test methods on a standard set of benchmark systems for informative comparisons. \paragraph{Direct tests of protein-ligand binding calculations} -Although it is still very difficult to convincingly verify convergence of many protein-ligand binding calculations, it is still important to compare the performance of various methods in real-world challenges. -Appropriate soft benchmarks are likely to be cases which are still relatively tractable, involving small proteins and simple binding sites. -We need a series of benchmark protein-ligand systems that introduce various challenges in a well-understood manner. -Systems should introduce none, one, two, or $N$ of the following challenges in various combinations: +Although it is still very difficult to convincingly verify convergence of many protein-ligand binding calculations, it is still important to compare the performance of various methods for these real-world challenges. +Appropriate soft benchmarks are likely to be cases which are still not overly challenging, involving small proteins and simple binding sites. +We propose defining a series of benchmark protein-ligand systems that systematically introduce specific challenges. +In particular, they should exemplify none, one, two, or $N$ of the following challenges, in various combinations: \begin{enumerate} \item Sampling challenges \begin{enumerate} \item Sidechains in the binding site rearrange on binding different ligands - \item Modest receptor conformational changes, such as loop motion + \item Modest backbone conformational changes, such as loop motion \item Large scale conformational changes, such as domain motions and allostery \item Ligand binding modes change unpredictably with small chemical modifications \item High occupancy water sites rearrange depending on bound ligand @@ -235,7 +249,7 @@ \subsubsection{Soft benchmarks} \item System challenges \begin{enumerate} \item Protonation state of ligand and/or protein changes on binding - \item Multiple protonation states of the ligand and/or receptor are relevant + \item Multiple protonation states of the ligand and/or receptor are relevant, due to pKas near the experimental pH, or the presence of multiple relevant tautomers \item Results are sensitive to buffer, salts or other environmental factors \end{enumerate} \item Force field challenges @@ -250,29 +264,31 @@ \subsubsection{Soft benchmarks} Eventually, some will become sufficiently well characterized and sampled that they become hard benchmarks. \subsection{Applications and limitations of benchmark systems} -Standard benchmark systems along the lines sketched above will allow potential solutions to be tested in a straightforward, reproducible manner. +Benchmark systems along the lines sketched above will allow potential advances in computational methods to be tested in a straightforward, reproducible manner. For example, force fields may be assessed by swapping new parameters, or even a new functional form, into an existing workflow to see the impact on accuracy for a hard benchmark test. Sampling methods may be assessed by using various enhanced sampling methods for either hard or soft sampling benchmarks, here without focusing on accuracy relative to experiment. -And system preparation tools could be varied to see how different approaches to assigning protonation states, modeling missing loops, or setting initial ligand poses, affect agreement with experiment---with the understanding that force field and sampling also play a role. -Such studies will be greatly facilitated by well-characterized standard benchmarks. +And system preparation tools could be varied to see how different approaches to assigning protonation states, modeling missing loops, or setting initial ligand poses, affect agreement of receptor-ligand binding calculations with experiment---with the understanding that force field and sampling also play a role. +Finally, comparisons across methods will be greatly facilitated by community acceptance of a set of standard cases: well-characterized and studied benchmarks utilized by the majority of developers and research groups, ideally on a routine basis. At the same time, there is a possibility that that some methods will inadvertently end up tuned specifically to generate good results for the set of accepted benchmarks. In such cases, the results for systems outside the benchmark set might still be disappointing. This means the field will need to work together to develop a truly representative set of benchmarks. This potential problem can also be mitigated by sharing of methods to enable broader testing by non-developers, and by participation in blinded prediction challenges, such as SAMPL and D3R, which confront methods with entirely new challenge cases. -\section{BENCHMARK SYSTEMS FOR BINDING PREDICTION} +\section{CURRENT BENCHMARK SYSTEMS FOR BINDING PREDICTIONS} \label{benchmarks} -No molecular systems have been explicitly accepted by the field as benchmarks for free energy calculations, but certain host molecules (see below) and designed binding sites in the enzyme T4 lysozyme have emerged as particularly helpful and widely studied test cases. +No molecular systems have yet been designated by the field as benchmarks for free energy calculations, but certain host molecules and designed binding sites in the enzyme T4 lysozyme have emerged as particularly helpful and widely studied test cases. Here, we describe these artificial receptors and propose specific host-guest and T4 lysozyme-ligand combinations as initial benchmark systems for free energy calculations. -We also point to several additional hosts and small proteins that also have potential to generate useful benchmarks in the future (Section~\ref{sec:updates}). -The present focus is on cases where experimental data are available and add value, rather than ones chosen specifically to test conformational sampling methods, where experimental data are not required (Section~\ref{subsec:benchmarktypes}). +We also point to several additional hosts and small proteins that also have potential to generate useful benchmarks in the future (Section~\ref{sec:futuresystems}). +We focus on cases where experimental data are available and add value, rather than cases chosen to test conformational sampling methods, for which experimental data are not required (Section~\ref{subsec:benchmarktypes}). \subsection{Host-guest benchmarks} +\label{sec:hgbenchmarks} Chemical hosts are small molecules, often comprising fewer than 100 non-hydrogen atoms, with a cavity or cleft that allows them to bind other compounds, called guests, with significant affinity. Hosts bind their guests via the same basic forces that proteins used to bind their ligands, so they can serve as simple test systems for computational models of noncovalent binding. Moreover, their small size, and, in many cases, their rigidity, can make it feasible to sample all relevant conformations, making for ``hard" benchmarks as defined above (Section~\ref{subsec:benchmarktypes}). -Furthermore, experiments can often be run under conditions that make the protonation states of the host and guest unambiguous. Under these conditions, the level of agreement of correctly executed calculations with experiment effectively reports on the validity of the force field (Section~\ref{pgph:accuracy}. +Furthermore, experiments can often be run under conditions that make the protonation states of the host and guest unambiguous. +Under these conditions, the level of agreement of correctly executed calculations with experiment effectively reports on the validity of the force field (Section~\ref{pgph:accuracy}). For a number of host-guest systems, the use of isothermal titration calorimetry (ITC) to characterize binding provides both binding free energies and binding enthalpies. Binding enthalpies can often also be computed to good numerical precision~\cite{henriksen_computational_2015}, so they provide an additional check of the validity of simulations. A variety of curated host-guest binding data is available on BindingDB at \url{http://bindingdb.org/bind/HostGuest.jsp}. @@ -280,7 +296,8 @@ \subsection{Host-guest benchmarks} For example, all members of the cyclodextrin family are chiral rings of glucose monomers; family members then differ in the number of monomers and in the presence or absence of various chemical substituents. For tests of computational methods ultimately aimed at predicting protein-ligand binding affinities in aqueous solution, water soluble hosts are, arguably, most relevant. On the other hand, host-guest systems in organic solvents may usefully test how well force fields work in the nonaqueous environment within a lipid membrane. -Here, we focus on two host families, the cucurbiturils \cite{freeman_cucurbituril_1981,mock_host-guest_1983}; and the octa-acids (more generally, Gibb deep cavity cavitands)~\cite{gibb_well-defined_2004, hillyer_synthesis_2016}, which have already been the subject of concerted attention from the simulation community, due in part to their use in the SAMPL blinded prediction challenges~\cite{muddana_sampl3_2012, muddana_sampl4_2014, yin_overview_2016}. +Here, we focus on two host families, the cucurbiturils \cite{freeman_cucurbituril_1981,mock_host-guest_1983} and the octa-acids (more generally, Gibb deep cavity cavitands)~\cite{gibb_well-defined_2004, hillyer_synthesis_2016}. +These have already been the subject of concerted attention from the simulation community, due in part to their use in the SAMPL blinded prediction challenges~\cite{muddana_sampl3_2012, muddana_sampl4_2014, yin_overview_2016}. \begin{figure*} \includegraphics[width=\textwidth]{figures/hosts.pdf} @@ -288,15 +305,16 @@ \subsection{Host-guest benchmarks} \end{figure*} \subsubsection{Cucurbiturils} - +\label{sec:cb} The cucurbiturils ({\bf Figure~\ref{hosts}}) are achiral rings of glycoluril monomers~\cite{freeman_cucurbituril_1981}. The first characterized family member, cucurbit[6]uril, has six glycoluril units, and subsequent synthetic efforts led to the five-, seven-, eight- and ten-monomer versions, cucurbit[n]uril (n=5,6,7,8,10)~\cite{liu_cucurbituril_2005}, which have been characterized to different extents. -Of note, the n=6,7,8 variants accommodate guests of progressively larger size, but are consistent in preferring to bind guests with a hydrophobic core sized to fit snugly into the relatively nonpolar binding cavity, along with at least one cationic moiety (though neutral compounds do bind~\cite{wyman_cucurbituril_2008, lee_deciphering_2015}) that forms stabilizing interactions with the oxygens of the carbonyl groups fringing both portals of the host~\cite{liu_cucurbituril_2005}. -Although derivatives of these parent compounds, have been made \cite{Lee:2003:AccountsofChemicalResearch, vinciguerra_synthesis_2015, assaf_cucurbiturils:_2015, Cong:2016:Org.Biomol.Chem.}, -most of the binding data published for this class of hosts pertain to the non-derivatized forms. A fairly extensive set of data is available in BindingDB at \url{http://bindingdb.org/bind/HostGuest.jsp}. +The n=6,7,8 variants accommodate guests of progressively larger size, but are consistent in preferring to bind guests with a hydrophobic core sized to fit snugly into the relatively nonpolar binding cavity, along with at least one cationic moiety (though neutral compounds do bind~\cite{wyman_cucurbituril_2008, lee_deciphering_2015}) that forms stabilizing interactions with the oxygens of the carbonyl groups fringing both portals of the host~\cite{liu_cucurbituril_2005}. +Although derivatives of these host molecules have been made \cite{Lee:2003:AccountsofChemicalResearch, vinciguerra_synthesis_2015, assaf_cucurbiturils:_2015, Cong:2016:Org.Biomol.Chem.}, +most of the binding data published for this class of hosts pertain to the non-derivatized forms. +A fairly extensive set of data is available in BindingDB at \url{http://bindingdb.org/bind/HostGuest.jsp}. -We propose cucurbit[7]uril (CB7) as the basis of one series of host-guest benchmark systems ({\bf Figure~\ref{hosts}}, {\bf Tables~\ref{cb7_benchmark1} and~\ref{cb7_benchmark2}}). +We propose cucurbit[7]uril (CB7) as the basis of one series of host-guest benchmark systems ({\bf Figure~\ref{hosts}}). This host is convenient experimentally, because it is reasonably soluble in water; and computationally, because it is quite rigid and lacks acidic or basic groups. In addition, it has attracted particular interest because of the high binding affinities of some guests, exceeding even the tightest-binding protein-ligand systems~\cite{liu_cucurbituril_2005, rekharsky_synthetic_2007, moghaddam_hostguest_2009, cao_attomolar_2014}. Finally, CB7 is already familiar to a number of computational chemistry groups, as it figured in two of the three SAMPL challenges that included host-guest components~\cite{muddana_sampl3_2012, muddana_sampl4_2014}, and it is currently the focus of the ``hydrophobe challenge''~\cite{schreiner_theoretical_2016}. @@ -320,7 +338,7 @@ \subsubsection{Cucurbiturils} For CB7, we have selected two sets of guests that were studied experimentally under uniform conditions (50 mM sodium acetate buffer, pH 4.74, 298K) by one research group ~\cite{liu_cucurbituril_2005, cao_attomolar_2014}. Each series is based on a common chemical scaffold, making it amenable to not only absolute but also alchemical relative free energy calculations (Section~\ref{sec:FEMethods}). One set is based on an adamantane core ({\bf Table~\ref{cb7_benchmark1}}), and the other on an aromatic ring ({\bf Table~\ref{cb7_benchmark2}}). These systems can be run to convergence to allow detailed comparisons among methods and with experiment. -Their binding free energies range from -5.99 to -17.19 kcal/mol, with the adamantane series spanning a particularly large range of free energies. +Their measured binding free energies range from -5.99 to -17.19 kcal/mol, with the adamantane series spanning a particularly large range. \paragraph{Prior studies provide additional insight into CB7's challenges} @@ -328,24 +346,30 @@ \subsubsection{Cucurbiturils} Sampling of the host appears relatively straightforward in CB7 as it is quite rigid and its symmetry provides for clever convergence checks~\cite{henriksen_computational_2015, monroe_converging_2014}. Due to its top-bottom symmetry, flips of guests from ``head-in'' to ``head-out'' configurations are not necessary to obtain convergence~\cite{fenley_bridging_2014}. However, sampling of the guest geometry can be a challenge, with transitions between binding modes as slow as 0.07 flips/ns~\cite{monroe_converging_2014}, and flexible guests also presenting challenges~\cite{monroe_converging_2014}. -As noted above, water sampling can also be an issue, with wetting/dewetting transitions occurring on the 50 ns timescale~\cite{rogers_role_2013}. +As noted above, water sampling can also be an issue, with wetting/dewetting transitions occurring on the 50 ns timescale~\cite{rogers_role_2013} when the guest is partly decoupled from the aqueous host in alchemical calculations. %System Salt and buffer conditions are also key. In addition to the strong salt-dependence of binding~\cite{moghaddam_new_2011}, acetic acid (such as in a sodium acetate buffer) can compete with guests for the binding site~\cite{moghaddam_hostguest_2009}. This may partially explain systematic errors in some computational studies~\cite{muddana_blind_2014,hsiao_prediction_2014}. Indeed, the difference between 50 mM sodium acetate buffer and 100 mM sodium phosphate buffer impacts measured binding free energies by 2.5-2.8 kcal/mol~\cite{muddana_blind_2014, muddana_sampl4_2014}. -Cationic guests could also have substantial and differing interactions with the counterions in solution as well, potentially lowering affinity relative to zero-salt conditions~\cite{muddana_sampl4_2014}. -Thus, one group found a 6.4-6.8 kcal/mol dependence on salt concentration~\cite{hsiao_prediction_2014}, possibly impacting other studies as well~\cite{monroe_converging_2014} +Cationic guests could also have substantial and differing interactions with the counterions in solution, potentially lowering affinity relative to zero-salt conditions~\cite{muddana_sampl4_2014}. +Additionally, CB7 can also bind cations fairly strongly~\cite{Isaacs:2009:Chem.Commun., cao_absolute_2014, Marquez:2004:J.Am.Chem.Soc.}. +Thus, one group found a 6.4-6.8 kcal/mol dependence on salt concentration~\cite{hsiao_prediction_2014} (possibly due to cation competition for the binding site), possibly impacting other studies as well~\cite{monroe_converging_2014}. + %Force field -Despite these issues, CB7 appears to be at the point where careful studies can probe the true accuracy of our force fields~\cite{henriksen_computational_2015, gao_binding_2015, yin_toward_2015}, and the results can be sobering, with RMS errors in the binding free energies as high as 8 kcal/mol~\cite{henriksen_computational_2015, monroe_converging_2014}. More encouragingly, the values of $R^2$ values can be as high as 0.92~\cite{henriksen_computational_2015}. -Some force fields appear relatively worse than others~\cite{hsiao_prediction_2014, muddana_prediction_2012}. +Despite these issues, CB7 appears to be at the point where careful studies can probe the true accuracy of our force fields~\cite{hsiao_prediction_2014, muddana_prediction_2012,henriksen_computational_2015, gao_binding_2015, yin_toward_2015}, and the results can be sobering, with RMS errors in the binding free energies as high as 8 kcal/mol~\cite{henriksen_computational_2015, monroe_converging_2014}. +More encouragingly, the values of $R^2$ values can be as high as 0.92~\cite{henriksen_computational_2015}. Calculated values are in many cases quite sensitive to details of force field parameters~\cite{monroe_converging_2014, moghaddam_new_2011, muddana_prediction_2012}. For example, modest modification of some Lennard-Jones parameters yielded dramatic improvements in calculated values~\cite{yin_toward_2015}, and host-guest binding data has, accordingly, been suggested as an input for force field development~\cite{yin_toward_2015, henriksen_computational_2015, gao_binding_2015}. Water structure around CB7 and calculated binding enthalpies also appear particularly sensitive to the choice of water model~\cite{rogers_role_2013, fenley_bridging_2014, gao_binding_2015}, and water is clearly important for modulating binding~\cite{nguyen_grid_2012}. The water model also impacts the number of sodium ions which must be displaced (in sodium-based buffer) on binding~\cite{gao_binding_2015, henriksen_computational_2015}. -Despite its apparent simplicity, CB7 is still a challenging benchmark that can put important issues into high relief. For example, in SAMPL4, free energy methods yielded $R^2$ values from 0.1 to 0.8 and RMS errors of about 1.9 to 4.9 kcal/mol for the same set of CB7 cases. This spread of results across rather similar methods highlights the need for shared benchmarks. Potential explanations include convergence difficulties, subtle methodological differences, and details of how the methods were applied~\cite{muddana_sampl4_2014}. Until the origin of such discrepancies is clear, it is difficult to know how accurate our methods truly are. +In summary, CB7 is still a challenging benchmark that can put important issues into high relief. +For example, in SAMPL4, free energy methods yielded $R^2$ values from 0.1 to 0.8 and RMS errors of about 1.9 to 4.9 kcal/mol for the same set of CB7 cases~\cite{muddana_sampl4_2014}. +This spread of results across rather similar methods highlights the need for shared benchmarks. +Potential explanations include convergence difficulties, subtle methodological differences, and details of how the methods were applied. +Until the origin of such discrepancies is clear, it is difficult to know how accurate our methods truly are. \begingroup \squeezetable @@ -407,7 +431,7 @@ \subsubsection{Gibb Deep Cavity Cavitands (GDCC)} (Note that OA and TEMOA have also been called OAH and OAMe, respectively~\cite{yin_overview_2016}.) Additional family members with other substituents around the portal have been reported, as has a new series in which the eponymic carboxylic groups are replaced by various other groups, including a number of basic amines~\cite{hillyer_synthesis_2016}. However, we are not aware of binding data for these derivatives. -In view of these other hosts, however, we propose the more general name Gibb deep cavity cavitands (GDCCs) for this family of hosts. +Because these closely related hosts are clearly in the same family but do not have eight acidic groups, and in recognition of the family's developer, we propose the more general name Gibb deep cavity cavitands (GDCCs) for this family of hosts. The binding cavities of the GDCCs are fairly rigid, though less so than the cucurbiturils. Some simulators report ``breathing'' motions that vary the diameter of the entry by up to 8~\AA \cite{mikulskis_free-energy_2014}; and, in some studies, the benzoic acid ``flaps'' around the entry occasionally flip upward and into contact with the guest~\cite{yin_sampl5_2016, tofoleanu_absolute_2016}, though this motion has not been verified experimentally. Additionally, the four priopionate groups protruding into solution from the exterior base of the cavity are all flexible. @@ -424,13 +448,18 @@ \subsubsection{Gibb Deep Cavity Cavitands (GDCC)} Second, elongated guests can generate ternary complexes, in which two OA hosts encapsulate one guest, especially if both ends of the guest are not very polar~\cite{gibb_guests_2009}. \paragraph{The proposed GDCC benchmark sets are drawn from SAMPL} -As a core benchmark series for this family, we propose two sets which formed part of the SAMPL4 and SAMPL5 challenges, based on adamantane derivatives (Table~\ref{gdcc_benchmark1}) and cyclic (aromatic and saturated) carboxylic acids (Table~\ref{gdcc_benchmark2}) binding to hosts OA and TEMOA with free energies of -3.7 to -7.6 kcal/mol. These cases offer aqueous binding data with a reasonably broad range of binding free energies, frequently along with binding enthalpies; the hosts and many or all of their guests are small and rigid enough to allow convincing convergence of binding thermodynamics with readily feasible simulations; and, like the cucurbiturils, they are already emerging as \emph{de facto} computational benchmarks, due to their use in the SAMPL4 and SAMPL5 challenges~\cite{muddana_sampl4_2014, yin_overview_2016}. +We propose the establishment of two GDCC benchmark sets, based on data which formed part of the SAMPL4 and SAMPL5 challenges. +One set is based on experiments carried out in phosphate buffer at pH 11.5, and the other on experiments in tetraborate buffer at pH 9.2 +The guests in the first set (Table~\ref{gdcc_benchmark1}) are adamantane derivatives and cyclic (aromatic and saturated) carboxylic acids (Table~\ref{gdcc_benchmark1}), which bind to hosts OA and TEMOA with free energies of -3.7 to -7.6 kcal/mol. +The second set of guests (Table~\ref{gdcc_benchmark2}) comprises carboxylic acids based on phenyl and cyclohexane cores. +Both sets offer aqueous binding data with free energies spanning about 4 kcal/mol, frequently along with binding enthalpies. +The hosts and many or all of their guests are small and rigid enough to allow convincing convergence of binding thermodynamics with readily feasible simulations; and, like the cucurbiturils, they are already emerging as \emph{de facto} computational benchmarks, due to their use in the SAMPL4 and SAMPL5 challenges~\cite{muddana_sampl4_2014, yin_overview_2016}. -\paragraph{OA introduces new challenges beyond CB7} Issues deserving attention when interpreting the experimental data and calculating the binding thermodynamics of these systems include the following: +\paragraph{The GDCC hosts introduce new challenges beyond CB7} Issues deserving attention when interpreting the experimental data and calculating the binding thermodynamics of these systems include the following: \begin{enumerate} \item{{\bf Tight exit portal}: The methyl groups of the TEMOA variant narrow the entryway and can generate a barrier to the entry or exit of guest molecules with bulky hydrophobic cores, though the degree of constriction is not as marked as for CB7 (above). -The TEMOA methyls groups can additionally hinder sampling of guest poses in the bound state, leading to convergence problems~\cite{yin_overview_2016} specific to TEMOA. } +The TEMOA methyl groups can additionally hinder sampling of guest poses in the bound state, leading to convergence problems~\cite{yin_overview_2016}} \item{{\bf Host conformational sampling}: Although the flexible propionate groups are not proximal to the binding cavity, they are charged and so can have long-ranged interactions. As a consequence, it may be important to ensure their conformations are well sampled, though motions may be slow~\cite{mikulskis_free-energy_2014}. @@ -453,31 +482,31 @@ \subsubsection{Gibb Deep Cavity Cavitands (GDCC)} \begin{table*} % \tabcolsep7.5pt \caption{Proposed GDCC Set 1 benchmark data} + \label{gdcc_benchmark1} -\begin{tabular}{lllp{1.5cm}>{\ttfamily}lS[table-format=-1.3, table-figures-uncertainty=1]S[table-format=-1.2, table-figures-uncertainty=1]} +\begin{tabular}{llllp{1.5cm}>{\ttfamily}lS[table-format=-1.3, table-figures-uncertainty=1]S[table-format=-1.2, table-figures-uncertainty=1]} \toprule -\multicolumn{1}{c}{ID$^{\rm a}$} & \multicolumn{1}{c}{name} & \multicolumn{1}{c}{PC CID$^{\rm b}$} & \multicolumn{1}{c}{2D} & \multicolumn{1}{c}{\rmfamily SMILES} & \multicolumn{1}{c}{$\Delta G^{\rm c}$ \,{(}kcal/mol{)}} & \multicolumn{1}{c}{$\Delta H^{\rm d}$ \,{(}kcal/mol{)}} \\ +\multicolumn{1}{c}{ID$^{\rm a}$} & \multicolumn{1}{c}{SAMPL$^{\rm b}$} & \multicolumn{1}{c}{name} & \multicolumn{1}{c}{PC CID$^{\rm c}$} & \multicolumn{1}{c}{2D} & \multicolumn{1}{c}{\rmfamily SMILES} & \multicolumn{1}{c}{$\Delta G^{\rm c}$ \,{(}kcal/mol{)}} & \multicolumn{1}{c}{$\Delta H^{\rm d}$ \,{(}kcal/mol{)}} \\ + \midrule -\multicolumn{7}{c}{Octa Acid binders} \\ +\multicolumn{8}{c}{OA host} \\ \midrule -3 / OA-G1 & 5-hexynoic acid & 143036 & \parbox[c]{1em}{\includegraphics[scale=0.15]{figures/143036.pdf}} & C\#CCCCC(=O)O & -5.40 \pm 0.003 & -7.71 \pm 0.05 \\ -4 / OA-G6 & 3-nitrobenzoic acid & 8497 & \parbox[c]{1em}{\includegraphics[scale=0.15]{figures/8497.pdf}} & C1=CC(=CC(=C1)[N+](=O)[O-])C(=O)O & -5.34 \pm 0.005 & -5.67 \pm 0.01 \\ -5 / OA-G2 & 4-cyanobenzoic acid & 12087 & \parbox[c]{1em}{\includegraphics[scale=0.15]{figures/12087.pdf}} & C1=CC(=CC=C1C\#N)C(=O)O & -4.73 \pm 0.01 & -4.45 \pm 0.08 \\ -6 / OA-G4 & 4-bromoadamantane-1-carboxylic acid & 12598766 & \parbox[c]{1em}{\includegraphics[scale=0.15]{figures/12598766.pdf}} & C1C2CC3CC(C2)(CC1C3Br)C(=O)O & -9.37 \pm 0.01 & -14.78 \pm 0.02 \\ -7 / OA-G3 & N,N,N-trimethylhexan-1-aminium & 84774 & \parbox[c]{1em}{\includegraphics[scale=0.15]{figures/84774.pdf}} & CCCCCC[N+](C)(C)C & -4.49 \pm 0.01 & -5.91 \pm 0.10 \\ -8 / OA-G5 & trimethylphenethylaminium & 14108 & \parbox[c]{1em}{\includegraphics[scale=0.15]{figures/14108.pdf}} & C[N+](C)(C)CCC1=CC=CC=C1 & -3.72 \pm 0.01 & -9.96 \pm 0.11 \\ +3 & OA-G1 & 5-hexynoic acid & 143036 & \parbox[c]{1em}{\includegraphics[scale=0.15]{figures/143036.pdf}} & C\#CCCCC(=O)O & -5.40 \pm 0.003 & -7.71 \pm 0.05 \\ +4 & OA-G6 & 3-nitrobenzoic acid & 8497 & \parbox[c]{1em}{\includegraphics[scale=0.15]{figures/8497.pdf}} & C1=CC(=CC(=C1)[N+](=O)[O-])C(=O)O & -5.34 \pm 0.005 & -5.67 \pm 0.01 \\ +5 & OA-G2 & 4-cyanobenzoic acid & 12087 & \parbox[c]{1em}{\includegraphics[scale=0.15]{figures/12087.pdf}} & C1=CC(=CC=C1C\#N)C(=O)O & -4.73 \pm 0.01 & -4.45 \pm 0.08 \\ +6 & OA-G4 & 4-bromoadamantane-1-carboxylic acid & 12598766 & \parbox[c]{1em}{\includegraphics[scale=0.15]{figures/12598766.pdf}} & C1C2CC3CC(C2)(CC1C3Br)C(=O)O & -9.37 \pm 0.01 & -14.78 \pm 0.02 \\ +7 & OA-G3 & N,N,N-trimethylhexan-1-aminium & 84774 & \parbox[c]{1em}{\includegraphics[scale=0.15]{figures/84774.pdf}} & CCCCCC[N+](C)(C)C & -4.49 \pm 0.01 & -5.91 \pm 0.10 \\ +8 & OA-G5 & trimethylphenethylaminium & 14108 & \parbox[c]{1em}{\includegraphics[scale=0.15]{figures/14108.pdf}} & C[N+](C)(C)CCC1=CC=CC=C1 & -3.72 \pm 0.01 & -9.96 \pm 0.11 \\ \midrule -\multicolumn{7}{c}{TEMOA/OAMe binders} \\ +\multicolumn{8}{c}{TEMOA/OAMe host} \\ \midrule -3 / OA-G1 & 5-hexynoic acid & 143036 & \parbox[c]{1em}{\includegraphics[scale=0.15]{figures/143036.pdf}} & C\#CCCCC(=O)O & -5.476 \pm 0.006 & -9.961 \pm 0.006 \\ -4 / OA-G6 & 3-nitrobenzoic acid & 8497 & \parbox[c]{1em}{\includegraphics[scale=0.15]{figures/8497.pdf}} & C1=CC(=CC(=C1)[N+](=O)[O-])C(=O)O & -4.52 \pm 0.02 & -9.1 \pm 0.1 \\ -5 / OA-G2 & 4-cyanobenzoic acid & 12087 & \parbox[c]{1em}{\includegraphics[scale=0.15]{figures/12087.pdf}} & C1=CC(=CC=C1C\#N)C(=O)O & -5.26 \pm 0.01 & -7.6 \pm 0.1 \\ -6 / OA-G4 & 4-bromoadamantane-1-carboxylic acid & 12598766 & \parbox[c]{1em}{\includegraphics[scale=0.15]{figures/12598766.pdf}} & C1C2CC3CC(C2)(CC1C3Br)C(=O)O & {ND$^{\rm e}$} & {ND$^{\rm e}$} \\ -7 / OA-G3 & N,N,N-trimethylhexan-1-aminium & 84774 & \parbox[c]{1em}{\includegraphics[scale=0.15]{figures/84774.pdf}} & CCCCCC[N+](C)(C)C & -5.73 \pm 0.06 & -6.62 \pm 0.2 \\ -8 / OA-G5 & trimethylphenethylaminium & 14108 & \parbox[c]{1em}{\includegraphics[scale=0.15]{figures/14108.pdf}} & C[N+](C)(C)CCC1=CC=CC=C1 & {ND$^{\rm e}$} & {ND$^{\rm e}$} \\ +3 & OA-G1 & 5-hexynoic acid & 143036 & \parbox[c]{1em}{\includegraphics[scale=0.15]{figures/143036.pdf}} & C\#CCCCC(=O)O & -5.476 \pm 0.006 & -9.961 \pm 0.006 \\ +4 & OA-G6 & 3-nitrobenzoic acid & 8497 & \parbox[c]{1em}{\includegraphics[scale=0.15]{figures/8497.pdf}} & C1=CC(=CC(=C1)[N+](=O)[O-])C(=O)O & -4.52 \pm 0.02 & -9.1 \pm 0.1 \\ +5 & OA-G2 & 4-cyanobenzoic acid & 12087 & \parbox[c]{1em}{\includegraphics[scale=0.15]{figures/12087.pdf}} & C1=CC(=CC=C1C\#N)C(=O)O & -5.26 \pm 0.01 & -7.6 \pm 0.1 \\ +7 & OA-G3 & N,N,N-trimethylhexan-1-aminium & 84774 & \parbox[c]{1em}{\includegraphics[scale=0.15]{figures/84774.pdf}} & CCCCCC[N+](C)(C)C & -5.73 \pm 0.06 & -6.62 \pm 0.2 \\ \bottomrule \end{tabular}\\ -$^{\rm a}$ Compound ID from ~\cite{sullivan_binding_2016} and SAMPL5 ID from ~\cite{yin_overview_2016}; $^{\rm b}$ PubChem Compound ID; $^{\rm c}$ Standard binding free energy from~\cite{sullivan_binding_2016}, where all measurements were done via ITC in 50 mM sodium phosphate buffer at pH 11.5 and 298 K. Uncertainties, drawn from the experimental paper, were computed from triplicate measurements taken with freshly made solutions of host and guest. However, based on personal communication with the authors, it may be advisable to regard the accuracy more conservatively, at $\sim$2\% for $\Delta$G and $\sim$6\% for $\Delta$H; $^{\rm d}$ measured binding enthalpy~\cite{sullivan_binding_2016}, subject to the same conditions/caveats as $^{\rm c}$. $^{\rm e}$ not done. +$^{\rm a}$ Compound ID from ~\cite{sullivan_binding_2016}; $^{\rm b}$ SAMPL5 ID from ~\cite{yin_overview_2016}; $^{\rm c}$ PubChem Compound ID; $^{\rm c}$ Standard binding free energy from~\cite{sullivan_binding_2016}, where all measurements were done via ITC in 50 mM sodium phosphate buffer at pH 11.5 and 298 K. Uncertainties, drawn from the experimental paper, were computed from triplicate measurements taken with freshly made solutions of host and guest. However, based on personal communication with the authors, it may be advisable to regard the accuracy more conservatively, at $\sim$2\% for $\Delta$G and $\sim$6\% for $\Delta$H; $^{\rm d}$ measured binding enthalpy~\cite{sullivan_binding_2016}, subject to the same conditions/caveats as $^{\rm c}$. $^{\rm e}$ not done. \end{table*} \endgroup @@ -485,7 +514,7 @@ \subsubsection{Gibb Deep Cavity Cavitands (GDCC)} \squeezetable \begin{table*} % \tabcolsep7.5pt -\caption{Proposed GDCC Set 2 benchmark data} +\caption{Proposed GDCC Set 2 benchmark data for binding to the OA host.} \label{gdcc_benchmark2} \begin{tabular}{lllp{2cm}>{\ttfamily}llS[table-format=-1.2, table-figures-uncertainty=1]} \toprule @@ -508,7 +537,7 @@ \subsubsection{Gibb Deep Cavity Cavitands (GDCC)} %SAMPLING As noted, two different host conformational sampling issues have been observed, with dihedral transitions for the proprionate groups occurring on 1-2 ns timescales~\cite{mikulskis_free-energy_2014}); motions of the benzoic acid flaps were also relatively slow~\cite{yin_sampl5_2016, tofoleanu_absolute_2016} though perhaps thermodynamically unimportant. -Guest sampling can also be an issue, at least in TEMOA~\cite{yin_overview_2016}, and this hosts's tight cavity may also have implications for binding entropy~\cite{yin_sampl5_2016}. +Guest sampling can also be an issue, at least in TEMOA~\cite{yin_overview_2016}, and this host's tight cavity may also have implications for binding entropy~\cite{yin_sampl5_2016}. % SYSTEM: Salt concentration strongly modulates binding affinity, at least for anions, and the nature of the salt also plays an important role~\cite{carnegie_anion_2014}. @@ -516,36 +545,39 @@ \subsubsection{Gibb Deep Cavity Cavitands (GDCC)} Some salts even bind to OA themselves, with perchlorate~\cite{gibb_anion_2011} and trichloroacetate~\cite{sokkalingam_binding_2016} being particularly potent, and thus will compete with guests for binding. Computationally, including additional salt beyond that needed for system neutralization changed binding free energies by up to 4 kcal/mol~\cite{tofoleanu_absolute_2016}. -Naively, protonation states of the guests might seem clear and unambiguous. -But since OA can bind guests of diverse net charges, the protonation state may not always be clear. -One study used absolute binding free energy calculations for different guest charge states, coupled with pKa calculations, and found that inclusion of pKa corrections and the possibility of alternate charge states of the guests affected calculated binding free energies by up to 2 kcal/mol~\cite{tofoleanu_absolute_2016}. As noted above, experimental evidence also indicates major pKa shifts on binding so that species such as acetate, formate and others would bind in neutral form at neutral pH~\cite{wang_itc_2016, sokkalingam_binding_2016}. -Even the host protonation state may be unclear; while OA is often assumed to have all eight carboxylic acids deprotonated at the basic pH of typical experiments, the four at the bottom are in close proximity, and these might make hydrogen bonds allowing retention of two protons~\cite{ewell_water_2008}. +Although the protonation states of the guests might seem clear and unambiguous, experimental evidence also indicates major pKa shifts on binding so that species such as acetate, formate and others could bind in neutral form at neutral pH~\cite{wang_itc_2016, sokkalingam_binding_2016}. +One study used absolute binding free energy calculations for different guest charge states, coupled with pKa calculations, and found that inclusion of pKa corrections and the possibility of alternate charge states of the guests affected calculated binding free energies by up to 2 kcal/mol~\cite{tofoleanu_absolute_2016}. +Even the host protonation state may be unclear. Although OA might be assumed to have all eight carboxylic acids deprotonated at the basic pH of typical experiments, the four at the bottom are in close proximity, and these might make hydrogen bonds allowing retention of two protons~\cite{ewell_water_2008}. Thus, there are uncertainties as to the host protonation state~\cite{muddana_sampl4_2014, ewell_water_2008}, which perhaps also could be modulated by guest binding. %FORCE FIELD, etc.: -%The challenges of system definition and conformational sampling appear to be greater for OA and TEMOA than for CB7, , so evidence of force field limitations, if present, is not yet as clear. There have been some comparisons of charge~\cite{mikulskis_free-energy_2014, muddana_sampl4_2014, monroe_converging_2014} and water models~\cite{yin_sampl5_2016}, but differences so far seem mostly inconclusive or at least not that large. Similarly, OPLS and GAFF results did not appear dramatically different in accuracy~\cite{bhakat_resolving_2016} -%The above paragraph could be cut for space if we need - -Several groups used different methods but the same force field and water model in SAMPL5, with rather varied levels of success because of discrepancies in calculated free energies~\cite{yin_overview_2016, bosisio_blinded_2016, bhakat_resolving_2016}. -However, some of these issues were resolved in follow-up work~\cite{bhakat_resolving_2016}, bringing the methods into fairly good agreement for the majority of cases~\cite{yin_sampl5_2016, bosisio_blinded_2016}. +The challenges of system definition and conformational sampling appear to be greater for OA and TEMOA than for CB7, as discussed above, so it is harder to draw definite conclusions regarding force field accuracy and differences among force fields. +Thus, although there have been some comparisons of charge~\cite{mikulskis_free-energy_2014, muddana_sampl4_2014, monroe_converging_2014} and water models~\cite{yin_sampl5_2016}, the resulting differences in computed binding thermodynamics so far seem mostly inconclusive or at least not that large. +Similarly, OPLS and GAFF results did not appear dramatically different in accuracy~\cite{bhakat_resolving_2016}. +It is worth noting that several groups using different computational approaches but the same force field and water model in SAMPL5 did not obtain identical binding free energies~\cite{yin_overview_2016, bosisio_blinded_2016, bhakat_resolving_2016}. +Some of these issues were resolved in follow-up work~\cite{bhakat_resolving_2016}, bringing the methods into fairly good agreement for the majority of cases~\cite{yin_sampl5_2016, bosisio_blinded_2016}. \subsection{Protein-ligand benchmarks: the T4 lysozyme model binding sites} - +\label{sec:t4} \begin{figure*} \includegraphics[width=\textwidth]{figures/lysozyme.pdf} -\caption{\label{fig:lysozyme} Benzene and hexylbenzene in the lysozyme L99A site, and phenol and 4,5,6,7-tetrahydroindole in the L99A/M102Q site (PDBs 4W52, 4W59, 1LI2, and 3HUA, respectively). The binding site shape is shown as a semi-transparent surface, and the protein shown with cartoons. In both cases, the structure with the smaller ligand is shown in green and that with the larger ligand is shown in blue, and the larger ligand induces a motion of helix F bordering the binding site. Phenol and 4,5,6,7-tetrahydroindole both also bind with an ordered water, though this does not occur for all ligands in the polar L99A/M102Q site.} +\caption{\label{fig:lysozyme} Benzene and hexylbenzene in the apolar lysozyme L99A site (left), and phenol and 4,5,6,7-tetrahydroindole in the polar L99A/M102Q site (right). +These structures are from PDBs 4W52, 4W59, 1LI2, and 3HUA, respectively. +The binding site cavities are shown as a semi-transparent surface, and the protein is shown with cartoons. +In both cases, the structure with the smaller ligand is shown in green and that with the larger ligand is shown in blue; the larger ligand (hexylbenzene in L99A, 4,5,6,7-tetrahydroindole in L99A/M102Q) induces a motion of helix F bordering the binding site. Phenol and 4,5,6,7-tetrahydroindole both also bind with an ordered water (red sphere, right panel), though this does not occur for all ligands in the polar L99A/M102Q site.} \end{figure*} Although we seek ultimately to predict binding in systems of direct pharmaceutical relevance, simpler protein-ligand systems can represent important stepping stones in this direction. Two model binding sites in T4 lysozyme have been particularly useful in this regard ({\bf Figure~\ref{fig:lysozyme}}). -These two binding sites, called L99A~\cite{morton_energetic_1995, morton_specificity_1995} and L99A/M102Q~\cite{wei_model_2002, graves_decoys_2005} for point mutations which create the cavities of interest, are created in artificial mutants of phage T4 lysozyme, and have been studied extensively experimentally and via modeling. +These two binding sites, called L99A~\cite{morton_energetic_1995, morton_specificity_1995} and L99A/M102Q~\cite{wei_model_2002, graves_decoys_2005} for point mutations which create the cavities of interest, have been studied extensively experimentally and via modeling. As protein-ligand systems, they introduce additional complexities beyond those observed in host-guest systems, yet they share some of the same simplicity. The ligands are generally small, neutral, and relatively rigid, with clear protonation states. -In many cases, substantial protein motions are absent, allowing calculated binding free energies to apparently converge relatively easily. -However, like host-guest systems, these binding sites are still surprisingly challenging~\cite{mobley_use_2006, mobley_confine_2007, mobley_predicting_2007, boyce_predicting_2009, jiang_free_2010, gallicchio_binding_2010, lim_sensitivity_2016}. In addition, precise convergence is sometimes difficult to achieve, and it is in all cases essentially impossible to fully verify. As a consequence, these are ``soft benchmarks" as defined above (Section~\ref{subsec:benchmarktypes}). -The importance of the lysozyme model sites is also driven by the relative wealth of experimental data. -It is relatively easy to identify new ligands and obtain high quality crystal structures and affinity measurements, allowing two different rounds of blind predictions testing free energy calculations~\cite{mobley_predicting_2007, boyce_predicting_2009}. +In many cases, substantial protein motions do not occur on binding, helping calculated binding free energies to reach apparent convergence relatively easily. +However, like host-guest systems, these binding sites are still surprisingly challenging~\cite{mobley_use_2006, mobley_confine_2007, mobley_predicting_2007, boyce_predicting_2009, jiang_free_2010, gallicchio_binding_2010, lim_sensitivity_2016}. +Thus, precise convergence is sometimes difficult to achieve, and it is in all cases essentially impossible to fully verify. As a consequence, these are ``soft benchmarks" as defined above (Section~\ref{subsec:benchmarktypes}). +The utility of the lysozyme model sites is also driven by the large body of available experimental data. +It has been relatively easy to identify new ligands and obtain high quality crystal structures and affinity measurements, and this has allowed two different rounds of blinded free energy prediction exercises~\cite{mobley_predicting_2007, boyce_predicting_2009}. \subsubsection{The apolar and polar cavities and their ligands} @@ -555,32 +587,31 @@ \subsubsection{The apolar and polar cavities and their ligands} One small downside of these binding sites is that the range of affinities is relatively narrow: about -4.5 to -6.7 kcal/mol in the apolar site~\cite{morton_energetic_1995, mobley_predicting_2007}, and about -4 to -5.5 kcal/mol in the polar site~\cite{boyce_predicting_2009}. Thus, even the strongest binders are not particularly strong, and the weakest binders tend to run up against their solubility limits. Still, these sites offer immensely useful tests for free energy calculations. -For both sites, fixed charge force fields seem to yield reasonably accurate free energies, with RMS errors between 1-2 kcal/mol, and some level of correlation with experiment, despite limited dynamic range~\cite{deng_calculation_2006, mobley_predicting_2007, boyce_predicting_2009, gallicchio_binding_2010, wang_identifying_2013}. +For both sites, fixed charge force fields seem to yield reasonably accurate free energies, with RMS errors of 1-2 kcal/mol, and some level of correlation with experiment, despite limited dynamic range~\cite{deng_calculation_2006, mobley_predicting_2007, boyce_predicting_2009, gallicchio_binding_2010, wang_identifying_2013}. System composition/preparation issues also do not seem to be a huge factor. Instead, sampling issues predominate: \begin{enumerate} -\item{Ligand binding mode/orientational sampling: The binding sites are buried and roughly oblong, with ligands which are similar in shape. +\item{Ligand orientation: These oblong binding sites are buried, and their ligands are similar in shape. Ligands with axial symmetry typically have at least two reasonably likely binding modes, but broken symmetry can drive up the number of likely binding modes. For example, phenol has two plausible binding modes in the polar cavity~\cite{graves_rescoring_2008, boyce_predicting_2009} but 3-chlorophenol has at least four, three of which appear to have some population in simulations~\cite{gallicchio_binding_2010}, because the chlorine could point in either direction within the site. Timescales for binding mode interconversion are relatively slow, with in-plane transitions on the 1-10 nanosecond timescale, and out-of-plane transitions (e.g. between toluene's two symmetry-equivalent binding modes) taking hundreds of nanoseconds (Mobley group, unpublished data).} -\item{Sidechain rearrangements: Some sidechains are known to reorganize when binding certain ligands. -The smallest ligands tend not to induce conformational changes, but larger ligands may induce sidechain rearrangements -- often, rotamer flips -- around the binding site region. -These can be slow in the tightly packed binding site. -This especially occurs for Val111 in the L99A site~\cite{morton_specificity_1995, mobley_confine_2007, jiang_free_2010} and Leu118, Val11, and Val103 in L99A/M102Q~\cite{wei_model_2002, wei_testing_2004, graves_rescoring_2008, boyce_predicting_2009}. -These sidechain motions typically present sampling problems for standard MD simulations~\cite{mobley_confine_2007, mobley_predicting_2007, boyce_predicting_2009, jiang_free_2010, wang_achieving_2012}. +\item{Sidechain rearrangement: Some sidechains reorganize when certain ligands, particularly larger ones, bind. +This is common in particular for the side chain of Val 111 in the L99A site~\cite{morton_specificity_1995, mobley_confine_2007, jiang_free_2010} and Leu 118, Val 111, and Val 103 in L99A/M102Q~\cite{wei_model_2002, wei_testing_2004, graves_rescoring_2008, boyce_predicting_2009}. +These sidechain motions can be slow, due to the tight packing of the binding site, and therefore can present sampling problems for standard MD simulations~\cite{mobley_confine_2007, mobley_predicting_2007, boyce_predicting_2009, jiang_free_2010, wang_achieving_2012}. } -\item{Backbone sampling: -Larger ligands induce shifts of the F helix, residues 107 or 108 to 115, adjacent to the binding site, allowing the site to enlarge. +\item{Backbone rearrangement: +Larger ligands induce shifts of the F helix (residues 107 or 108 to 115), which is adjacent to the binding site, allowing the site to enlarge. This occurs in both binding sites~\cite{wei_testing_2004, boyce_predicting_2009, merski_homologous_2015}, but is best characterized for L99A~\cite{merski_homologous_2015}. There, addition of a series of methyl groups from benzene up to n-hexylbenzene causes a conformational transition in the protein from closed to intermediate to open conformations. } \end{enumerate} -{\bf Tables~\ref{apolar_benchmark} and~\ref{polar_benchmark}} introduce proposed benchmark sets for the apolar and polar cavities, giving ligands potentially amenable to both absolute and relative free energy calculations, and spanning the range of available affinities. -Co-crystal structures are available in most cases, and the PDB IDs are provided in the tables. +{\bf Tables~\ref{apolar_benchmark} and~\ref{polar_benchmark}} introduce two initial benchmark sets based on the apolar and polar T4 lysozyme binding sites. +These sets include ligands amenable to both absolute and relative free energy calculations, and have affinities that cover the currently available experimental range. +Co-crystal structures are available in most cases; the corresponding PDB IDs are provided in the tables. The selected ligands span a range of challenges and levels of difficulty, ranging from fairly simple to including most of the challenges noted above. Essentially all of them have been included in at least one prior computational study, and some have appeared in a variety of prior studies. Additional known ligands and non-binders are available, with binding affinities available for 19 compounds in the L99A site~\cite{eriksson_cavity-containing_1992, morton_energetic_1995, mobley_predicting_2007} and 16 in L99A/M102Q~\cite{wei_model_2002, graves_decoys_2005, boyce_predicting_2009}. @@ -646,63 +677,214 @@ \subsubsection{The apolar and polar cavities and their ligands} \subsubsection{Computational challenges posed by the T4 lysozyme benchmarks} Early work on the lysozyme sites focused on the difficulty of predicting binding modes~\cite{mobley_use_2006, mobley_predicting_2007, boyce_predicting_2009} because of the slow interconversions noted above. -Docking methods often can generate reasonable poses spanning most of the important possibilities~\cite{mobley_use_2006, mobley_predicting_2007, boyce_predicting_2009, graves_rescoring_2008} but do not accurately predict the binding mode of individual compounds~\cite{mobley_predicting_2007, boyce_predicting_2009, graves_rescoring_2008}. -Thus it appears necessary to consider the possibility of multiple binding modes; this is also important since some ligands actually populate multiple binding modes~\cite{boyce_predicting_2009}. +Docking methods often can generate reasonable poses spanning most of the important possibilities~\cite{mobley_use_2006, mobley_predicting_2007, boyce_predicting_2009, graves_rescoring_2008} but do not accurately predict the binding modes of individual compounds~\cite{mobley_predicting_2007, boyce_predicting_2009, graves_rescoring_2008}. +Thus, binding calculations must explore multiple potential binding modes, especially as some ligands actually populate multiple poses at equilibrium~\cite{boyce_predicting_2009}. In a number of studies, candidate binding modes from docking are relaxed with MD simulations, then clustered to select binding modes for further study. It turns out an effective binding free energy for each distinct candidate binding mode can be computed separately~\cite{mobley_use_2006} and combined to find the population of each binding mode and determine the overall binding free energy. -However, this is costly since each candidate binding mode requires a full binding free energy calculation. - -Relative binding free energy calculations do not dramatically simplify the situation. -Introduction of a ligand modification can leave the binding mode uncertain (e.g., introducing a chlorine onto phenol leaves at least two possible binding modes even if the binding mode of phenol is known)~\cite{boyce_predicting_2009}. -A na{\" i}ve solution is to consider multiple possible binding modes in relative free energy calculations~\cite{boyce_predicting_2009}, but this generates multiple results; determining the true relative binding free energy requires additional information~\cite{mobley_perspective_2012}. Enhanced sampling approaches provide one possible solution to the binding mode problem. -Particularly, with $\lambda$ or Hamiltonian exchange techniques, ligands can easily switch between binding modes when they are non-interacting unless they are restrained, and then moves in $\lambda$ space can allow transitions back to the interacting state. +However, this is costly, since each binding mode requires a full binding free energy calculation. + +Relative binding free energy calculations do not dramatically simplify the situation,as +introduction of a ligand modification can leave the binding mode uncertain. +For example, adding a chlorine to phenol generates at least two binding mode variants, even if the binding mode of phenol is known)~\cite{boyce_predicting_2009}. +A na{\" i}ve solution is to consider multiple possible binding modes in relative free energy calculations~\cite{boyce_predicting_2009}, but this generates multiple results; determining the true relative binding free energy requires additional information~\cite{mobley_perspective_2012}. +Enhanced sampling approaches provide one possible solution to the binding mode problem. +For example, $\lambda$ or Hamiltonian exchange techniques can incorporate sampling on an artificial energy surface where the ligand does not interact with the protein, and thus can readily reorient. Thus, approaches employing this strategy can naturally sample multiple binding modes~\cite{gallicchio_binding_2010, wang_identifying_2013}. While sidechain sampling has been a significant challenge, it is possible to use biased sampling techniques such as umbrella sampling to deliberately compute and include free energies of sampling slow sidechain rearrangements~\cite{mobley_confine_2007}. -However, this is not a general solution, since it requires knowing what sidechains might rearrange on binding and then expending substantial computational power on sampling free energy landscapes for these rearrangements. -An apparently better general strategy is including sidechains in enhanced sampling regions selected for Hamiltonian exchange~\cite{jiang_free_2010, khavrutskii_improved_2011} or REST~\cite{wang_achieving_2012}, allowing sidechains to be alchemically softened or torsion barriers lowered (or both), to enhance sampling at alchemical intermediate states. -With swaps between $\lambda$ values, enhanced sidechain sampling at intermediate states can propagate to all states, improving convergence~\cite{jiang_free_2010, wang_achieving_2012}. +However, this is not a general solution, since it requires knowing what sidechains might rearrange on binding and then expending substantial computational power to map the free energy landscapes for these rearrangements. +An apparently better general strategy is including sidechains in enhanced sampling regions selected for Hamiltonian exchange~\cite{jiang_free_2010, khavrutskii_improved_2011} or REST2~\cite{wang_achieving_2012}, {\bf allowing sidechains to be alchemically softened or torsion barriers lowered (or both), to enhance sampling at alchemical intermediate states}. +%{\bf I don't understand the bolded text. For one thing, my impression is that REST only modifies T, and leaves the Hamiltonian unchanged.} +% DLM: REST2 -- which is what they currently use -- they like to describe as modifying T of a "local region" but really it's just modifying the Hamiltonian for selected residues, so it's a type of Hamiltonian exchange. You pick sidechains with interactions you want to soften, alchemically, and the softening is done at intermediate lambda values. How would you suggest clarifying the phrasing? (If my explanation is still unclear we could discuss on Slack or on the phone.) +Reduction of energy barriers at intermediate $\lambda$ states in this scheme allows enhanced conformational sampling at these $\lambda$ values. +Then, with swaps between $\lambda$ values, enhanced sidechain sampling at intermediate states can propagate to all states, improving convergence~\cite{jiang_free_2010, wang_achieving_2012}. Larger protein conformational changes in lysozyme have received less attention, partly because until very recently they seemed to be a peculiar oddity only rarely observed; i.e., for ligands 4,5,6,7-tetrahydroindole and benzyl acetate in the polar site~\cite{boyce_predicting_2009}. However, recent work noted above highlighted how a helix in the apolar cavity can open to accommodate larger ligands~\cite{merski_homologous_2015}. Timescales for this motion appear to be on the order of 50 ns, so it can pose sampling challenges, even for relative free energy calculations~\cite{lim_sensitivity_2016}. -Including part of the protein in the enhanced sampling region via REST2 provides some benefits, but sampling these motions will likely prove a valuable test for enhanced sampling methods. - -%This paragraph could be cut for space since it's speculative: -%Potentially, water sampling could also be a challenge in the polar binding site, as phenol binds with one ordered water~\cite{wei_model_2002} but to our knowledge, timescales for water sampling have not yet been examined, other than noting that water does not enter the polar binding site as ligands are removed~\cite{boyce_predicting_2009}. The apolar cavity is dry both with ligands present and with empty, so water sampling in the there cavity seems unlikely to be a challenge. - -\section{THE FUTURE OF BENCHMARKS AND OF THIS REVIEW} +Including part of the protein in the enhanced sampling region via REST2 (described above) provides some benefits~\cite{lim_sensitivity_2016}, but sampling these motions will likely prove a valuable test for enhanced sampling methods. + +Accounting for water exchange into and out of these buried binding sites could also be a challenge in the polar binding site, as phenol binds with one ordered water~\cite{wei_model_2002}. However, to our knowledge, the simulation timescales for water sampling have not yet been examined, other than noting that water does not enter the polar binding site as ligands are removed~\cite{boyce_predicting_2009}. +In contrast, the apolar cavity is dry, both with and without a bound ligand, so water sampling is unlikely to be a challenge in this case. + +\section{FUTURE BENCHMARK SYSTEMS} +\label{sec:futuresystems} +Although the benchmark systems detailed above are useful, more systems are needed to expand the dataset, broaden the range of challenges, and bridge to biomedically relevant protein-ligand binding. +The following subsections discuss additional systems that have already been used to test computational methods and that may be suited for development as new benchmark systems in the near future. + +\subsection{Host-Guest Systems} +The cyclodextrins (CDs; Section~\ref{sec:hgbenchmarks}) are a particularly promising source of additional host-guest benchmark sets. +Cyclic glucose polymers, the CDs are produced from starch by an enzymatic process, and are available in gram quantities at low cost from multiple suppliers. +Many experimental binding data, for varied guest molecules, are available for this class of host molecules, especially for 6-membered $\alpha$-CD and 7-membered $\beta$-CD, which both have adequate aqueous solubility. +Indeed, a thorough review 1998 review tabulates hundreds of binding data~\cite{rekharsky_complexation_1998}, and many additional measurements have been published since then; e.g. references~\cite{Connors:1997:Chem.Rev., Carrazana:2005:J.Phys.Chem.B, Cotner:1998:J.Org.Chem., Wszelaka-Rylik:2013:JThermAnalCalorim, Shu:2007:BritishJournalofPharmacology, Rodriguez-Perez:2006:J.Pharm.Sci., Mic:2013:AIPConferenceProceedings}. +A number of these studies were done calorimetrically, and thus provide not only binding free energies but also binding enthalpies. +Fewer data are available for 8-membered $\gamma$-CD, and the greater diameter of this host makes for a less well-defined and floppier binding cavity. +Overall, the amount of existing binding data for CDs greatly exceeds what is currently available for CB7 and other cucurbiturils. + +Structurally, the binding cavity of $\beta$-CD is about the same size as CB7 (Section~\ref{sec:cb}), but the cyclodextrins are more flexible than CB7, as their glucose monomers are joined by one single bond, whereas CB7's glycouril monomers are joined by two single bonds. +The CDs also appear easier to derivatize than the cucurbiturils. In particular, varied substituents may be appended by reactions involving secondary hydroxyls at the wide entry and primary hydroxyls at the narrow entry~\cite{Fromming:1994:CyclodextrinsinPharmacy, Dodziuk:2006:, Szente:1999:AdvancedDrugDeliveryReviews, Qu:2002:JournalofInclusionPhenomena, Jindrich:2005:J.Org.Chem.}, with the caveat that generating pure products can be difficult, because there are so many hydroxyls that may be modified. +Binding data on CD derivatives could be quite useful as a means of adding chemical diversity to host-guest benchmark sets, and such data are already available for some derivatives (see, e.g., \cite{rekharsky_complexation_1998, Faugeras:2012:Eur.J.Org.Chem.}). + +The CDs are computationally tractable \cite{Mark:1994:J.Am.Chem.Soc., Luzhkov:1999:ChemicalPhysicsLetters, Bea:2002:TheorChemAcc, Chen:2004:BiophysicalJournal, Sellner:2008:J.Phys.Chem.B, Cai:2009:J.Phys.Chem.B, Wickstrom:2013:J.Chem.TheoryComput., Shi:2014:TheorChemAcc, henriksen_computational_2015, Zhang:2015:J.Chem.TheoryComput., Khuntawee:2016:CarbohydratePolymers, Gebhardt:2016:FluidPhaseEquilibria, wickstrom_parameterization_2016}, and thus can be used for ``hard'' benchmark sets to test force fields (Section~\ref{pgph:accuracy}). +One complicating feature for calculations, relative to CB7, is that the two entryways to a CD are not equivalent, and asymmetric guest molecules may prefer to bind ``head-in'' or ``head-out''. +As these two binding modes typically do not interchange on the microsecond timescale, they must be considered separately when one computes binding thermodynamics ~\cite{henriksen_computational_2015}. +It is also worth noting that, because the CDs are glucose polymers, they may be best modeled with dedicated carbohydrate force fields \cite{Kirschner:2008:J.Comput.Chem., Cezard:2011:PhysicalChemistryChemicalPhysics, Guvench:2011:J.Chem.TheoryComput., Xiong:2015:CarbohydrateResearch}, rather than generalized small molecule force fields. +A number of CD (and other host guest) binding systems are available for download in electronic format at the BindingDB~\cite{Liu:2007:Nucl.AcidsRes., Gilson:2016:Nucl.AcidsRes.} website (\url{www.bindingdb.org/bind/HostGuest.jsp}) + +\subsection{Protein-Ligand Systems} +On the order of a million experimental protein-ligand binding measurements are currently accessible through open-access databases, notably BindingDB~\cite{Liu:2007:Nucl.AcidsRes.}, ChEMBL~\cite{Bento:2014:NuclAcidsRes} and PubChem~\cite{Kim:2016:NucleicAcidsRes.}. +These databases can be valuable sources of, or at least starting points for, new protein-ligand benchmark datasets. +In fact, automated procedures have already been used to extract about 700 downloadable validation sets (\url{www.bindingdb.org/validation\_sets/index.jsp}), each +comprising a congeneric series of \~10-50 ligands with binding data against a defined protein, and a cocrystal structure for at least one of the ligands in the series. +However, there is still a need for a smaller collection of highly optimized benchmark sets as research foci for the computational chemistry community. +Such sets should exemplify specific challenges not well covered by existing benchmark systems; be based upon high quality binding measurements for at least ~20 ligands, measured by consistent procedures across all ligands; and include crystal structures for the apo-protein and cocrystal structures for multiple ligands. +Analysis and extraction of such sets from the big databases is a promising future direction. +Here, however, we take the less systematic but more expedient approach of considering several systems that have already proven themselves to be computationally tractable and informative. + +\subsubsection{Constructed Binding Sites in Cytochrome C Peroxidase} +Two artificial binding sites have been designed into the enzyme cytochrome C peroxidase (CCP): the ``closed'' site, created by the mutation W191G~\cite{fitzgerald_ligand-gated_1996, rocklin_blind_2013}; and the ``open'' site mutant, created by supplementing mutation W191G with partial deletion of a loop, and thus opening the site to the outside of the protein~\cite{musah_artificial_2002, rosenfeld_excision_2002, rocklin_blind_2013}. +As for the artificial binding cavities in T4 lysozyme (Section~\ref{sec:t4}), the two CCP sites bind simple, fragment-like ligands with modest binding free energies (e.g., -3 to -7 kcal/mol for the open site \cite{rocklin_blind_2013}); discovery of new ligands is relatively straightforward~\cite{brenk_probing_2006, rocklin_blind_2013}; and new crystal structures can be obtained fairly readily. +The protein has relatively modest size (around 280 residues) as in, for example, the 1KXM structure. + +What is new is that the CCP binding sites contain an ASP which appears to be charged, at least in the presence of some ligands, as evidenced by observed interactions in structures, and by experiments where replacing the ASP with ASN abolishes binding of imidazoles, despite minimal changes to the binding site structure~\cite{Fitzgerald:1995:ProteinScience}. +This side-chain interacts at short range with the ligands, which may be cationic or polar neutral compounds. +As a consequence, the CCP sites challenge the ability of computational methods to accurately account for strong electrostatic interactions in the low-dielectric interior of a protein \cite{rocklin_blind_2013}. +Indeed, free energy calculations with two different force fields and distinct computational methods were found to overestimate the range of affinities, across a series of ligands, about three-fold, for the closed site~\cite{banba_free_2000, banba_efficient_2000}, and a similar pattern appears to hold for the open site~\cite{rocklin_blind_2013}. +Further analysis suggests that these overestimates stem from overestimated electrostatic interactions in these buried sites~\cite{rocklin_blind_2013}, perhaps because the force fields used did not account explicitly for electronic polarization. +Additionally, the fact that different CCP ligands have different net charges may pose methodological challenges for some free energy techniques~\cite{rocklin_calculating_2013}. +One modeling challenge is that CCP contains a heme, providing additional setup challenges (though since this is such an important cofactor, these may be worthwhile to face). +In summary, these model CCP binding sites appear challenging yet tractable, and have already yielded insight regarding possible directions for force field improvements. +Thus, they are good candidates to provide new benchmark sets in the near future. + + +\subsubsection{Thrombin} +Human thrombin, an enzyme of about 300 residues, is interesting both as a drug target related to blood coagulation and as a representative serine protease. +Binding and structural data are available for a wide variety of ligands (\cite{baum_non-additivity_2010, StefanicAnderluh:2005:J.Med.Chem., Ueno:2005:Bioorganic&MedicinalChemistryLetters, Putta:2005:J.Med.Chem., baum_non-additivity_2010} and others), +and have already been used as the basis for free energy simulations ~\cite{wang_achieving_2012, schrodinger_accurate_2015, calabro_elucidation_2016}. +Some of the experimental binding data, obtained calorimetrically, highlight interesting trends, such as non-additivity (positive coupling) between substitutions at different locations on the chemical scaffolds of two compound series~\cite{baum_non-additivity_2010}. + +One computational study obtained relative free energies for two compound series that largely captured experimental trends and achieved an overall mean unsigned error of 0.74 kcal/mol over a range of roughly 5 kcal/mol. +However, the accuracy of the results differed significantly between the two series~\cite{calabro_elucidation_2016}. +Another relative binding free energy study for thrombin found that the results changed, depending on the initial conformation of the ligands, and showed that enhanced sampling techniques could reduce the dependnce on starting conformation~\cite{wang_achieving_2012}. +These prior studies suggests thrombin may be at a ``sweet spot'' for benchmark systems, where the system is relatively tractable, and encouraging results have been obtained, but where there are still clear challenges. + + +\subsubsection{Bromodomain proteins} +Bromodomains (BRDs) are a family of protein domains of about 100 residues that bind actylated lysine residues at the surface of histones and thus read out epigenetic markers. +Bromodomains are present in many human proteins, and are being explored as potential drug targets for diseases including cancer and atherosclerosis~\cite{aldeghi_accurate_2016}. +These compact domains can be expressed, purified and studied as independent proteins, and are associated with a growing body of small molecule binding data \cite{Filippakopoulos:2010:Nature, Chung:2011:J.Med.Chem., Hewings:2012:J.Med.Chem., Bamborough:2015:J.Med.Chem., Ran:2015:J.Med.Chem., Zhang:2013:J.Med.Chem., Gosmini:2014:J.Med.Chem., Yang:2016:BioorganicChemistry}, including some obtained by isothermal titration calorimetry \cite{Filippakopoulos:2010:Nature, Chung:2011:J.Med.Chem., Hewings:2012:J.Med.Chem., Gosmini:2014:J.Med.Chem.}. +%DLM: Contact Aldeghi, etc. for updates (new cite forthcoming?) -- waiting for response + +The small size of bromodomains, and the fact that their binding sites are relatively solvent-exposed, makes them particularly suitable for free energy simulations. +One recent study, which used alchemical techniques to compute absolute binding free energies of 11 different ligands for bromodomain BRD4~\cite{aldeghi_accurate_2016}, achieved a remarkable level of accuracy, RMS error $0.8\pm0.2$ kcal/mol, for binding free energies spanning a range of \~4 kcal/mol. +Docking calculations included in the same study did not work as well. +The compounds studied were diverse, and therefore not amenable to relative free energy methods in which one ligand is computationally converted into another. +Overall, then, this appears to be class of systems that could yield relatively tractable and informative protein-ligand benchmark systems. +One known challenge is that some ligands have multiple plausible binding modes \cite{aldeghi_accurate_2016}. + In addition, a diverse ligand series can pose severe challenges for relative free energy techniques. + + +\subsubsection{Other protein-ligand systems} +Several other systems may be of possible interest because of the wealth and quality of experimental data, the extent of prior computational work, or the combination of pharmaceutical relevance with important challenges. +None of these seem to be well-characterized or well-studied enough yet to be benchmark systems, but they may be interesting choices for the future. +The first system we would put in this category is trypsin, and especially binding of benzamidine and its derivatives. +While trypsin has been the focus of several free energy studies~\cite{talhout_understanding_2003, villa_sampling_2003, jiao_calculation_2008, jiao_trypsin_2009, de_ruiter_efficient_2012} and the SAMPL3 challenge~\cite{Newman:2011:JComputAidedMolDes, skillman_sampl3_2012}, it also appears to be subject to extremely slow protein motions (on the tens of microsecond timescale) which are only beginning to be characterized~\cite{plattner_protein_2015}. +As a consequence, short free energy calculations may appear to converge~\cite{talhout_understanding_2003, jiao_calculation_2008, jiao_trypsin_2009}, but longer calculations can reveal a profound lack of convergence~\cite{plattner_protein_2015}. +Therefore, trypsin may become a good benchmark for studying binding that is coupled with slow conformational dynamics; enhanced sampling methods and +Markov state models \cite{plattner_protein_2015} may be particularly helpful here. + +HIV integrase is also of potential interest as a benchmark system. +Indeed, blind predictions for a set of fragments binding to several sites formed part of the basis of the SAMPL4 challenge~\cite{mobley_blind_2014, peat_interrogating_2014}. +Although the range of affinities in this set is modest (200 to 1450 $\mu M$~\cite{peat_interrogating_2014}), many crystal structures are available, as is a wealth of verified nonbinders~\cite{peat_interrogating_2014}. +One group actually did remarkably well using free energy calculations to recognize binding modes and predict nonbinders~\cite{gallicchio_virtual_2014, mobley_blind_2014}. +Coupled with other HIV integrase data available in the literature and binding databases for larger ligands or other series, this may make this system attractive for future benchmarks. + +The protein FKBP has also been the focus of several different free energy studies on the same series of ligands over the years~\cite{shirts_calculating_2004, fujitani_direct_2005, jayachandran_parallelized-over-parts_2006, lee_calculation_2006, wang_absolute_2006, fujitani_massively_2009, ytreberg_absolute_2009}. +However, in many cases, differences in calculated values between different studies -- even with what appears to be the same force field and system preparation -- are larger than differences relative to experiment~\cite{fujitani_direct_2005, jayachandran_parallelized-over-parts_2006}. +Some challenges are clear, such as conformational sampling for some of the larger ligands~\cite{shirts_calculating_2004}, and there have been some suggestions of force field issues and long equilibration times for the protein~\cite{fujitani_massively_2009}. +Perhaps this may be suitable as a benchmark system in the near-term as well. + +A number of other proteins also have strong potential to generate useful benchmark sets. For example, free energy calculations have been carried out for influenza neuraminidase inhibitors ~\cite{smith_dihydropyrancarboxamides_1998} with some success~\cite{michel_protein_2006}, but other series include more complicated structure-activity relationships and are associated with protein loop motions that may be difficult to model~\cite{kerry_structural_2013}. +Periplasmic oligopeptide binding protein A (OppA) binds a series of two to five-residue peptides, for which there exists a large amount of calorimetric and crystallographic binding data~\cite{tame_crystal_1995, davies_relating_1999, sleigh_crystallographic_1999}, and the system appears challenging but potentially tractable for free energy calculations~\cite{maurer_calculation_2016}. +The JNK kinase may pose an interesting conformational sampling challenge, due to its slow interconversion between binding modes for some ligands~\cite{kaus_how_2015}. +And the ongoing series of Drug Design Data Resource (D3R) \cite{} blinded challenges may also be source of informative protein-ligand systems (drugdesigndata.org/about/datasets) that are familiar to the computational chemistry community. +As noted above, however, many other protein-ligand binding systems have been characterized experimentally, and a systematic filtering would undoubtedly yield more benchmark candidates. + +\section{HOW TO USE BENCHMARK SYSTEMS} +Benchmark systems will have multiple uses, as discussed above, but not all benchmark systems can cover all uses. +Some will be particularly valuable for testing accuracy relative to experiment. +For this purpose, relatively large numbers of ligands are needed to afford meaningful statistics. +Other benchmark systems will be more useful for testing sampling techniques, and still others will, at least initially, serve as test beds to determine the sensitivity of computational results to various factors. + +In our view, benchmark systems will serve also to help design careful computational experiments. +For example, researchers can test whether a particular method is sampling the motions which others have already shown to be important, or how the choice of starting conformation impacts the rate of convergence to a known, gold standard value for a particular force field and system composition. +The availability of benchmark systems will also facilitate comparisons where only a single piece of a workflow is modified. +For example, one may ask how results change if a different protonation state assignment tool is used to prepare a protein. +Of course, such comparisons can already be done, but the results will be far more useful in the context of generally accepted and widely used test cases. + +We hope that, ultimately, results from reliable ``gold standard'' binding free energy computations will be available for a set of benchmark systems. +These would be from fully converged binding free energy calculations, and give correct results for a particular force field and system preparation, allowing quantitative comparison of the force field results with experiment. +Such results will also facilitate a great deal of science on method efficiency, as new methods which purport to be more efficient could easily and \emph{automatically} be run on a standard set of systems to see how much more efficient they are than the (perhaps brute-force) method which yielded the gold standard results. +Thus, various enhanced sampling methods could easily be observed to have strengths and weaknesses on known problem classes. +These systems will allow automated testing of the efficiency of new methods on real-world problems. + +\section{WE NEED WORKFLOW SCIENCE} +While the benchmark systems discussed here will already be useful, to fully realize their benefits a great deal of engineering needs to be done to facilitate workflow science. +Currently, a wide variety of computational tools are available for different stages of the free energy calculation process, from system preparation (protonation state assignment, building in missing residues and loops, adding counterions, etc.) to force field assignment, to planning and conducting the calculations themselves (choice of method, simulation package, and so on). +Often, each set of tools lives in its own ecosystem and is not designed to be easily interchangeable with tools from another ecosystem. +This makes it very difficult to systematically compare methods that differ only in one respect; instead, one must adopt an entirely different toolset to change one aspect of a procedure. +For example, swapping different tools for assigning protein protonation states could yield valuable insights into the relative merits of these tools and the importance of protonation at specific residues, but currently, this is, at best, an arduous task. + +\subsection{Workflow automation is needed} +At the most basic level, we need to allow calculations to be easily repeated on all of the benchmark systems via automated workflows. +One should not have to become an expert in the systems being studied in order to be able to successfully apply calculations to them; inputs should be easily available and repeating calculations should become fully automated so that a new method can be tested by simply specifying the set of benchmarks to run on. + +To achieve this, at least two major innovations are needed. +First, we need automated workflows that can proceed from the specification of a system to target to yielding the desired results without human intervention. +Second, we need a standard data structure for input to and output from these workflows so that people can easily obtain inputs for benchmark systems and only change the component they want to change (such as the force field or system preparation) and leave the other components unchanged so that, in an automated manner, they can focus their testing on only the components they want to test. + +\subsection{Analysis automation will also be needed} +At the most basic level, we can simply check whether we are getting the expected answer for each calculation performed, at least for systems where a gold standard result is available. +However, this does not provide nearly enough insight, especially in cases of failure, where we would like insight into \emph{why} we failed. +Are the relevant motions being sampled? +Do we have the right protonation states and binding modes? +Many other factors may need to be considered. +We need ways to automatically check that we are sampling the right motions, identifying correct binding modes/conformations, and so on, without having to become experts on the specific systems examined. +Probably we will need to define ways to automatically specify what order parameters should be monitored to assess for adequate sampling. + +\subsection{Modularization will be key} +To achieve these goals, researchers should develop or package tools so that they take a set of well specified inputs and provide well specified outputs in an interchangeable way. +This may involve containerizing key pieces of workflows such as in Docker~\cite{docker_what_2015} or Singularity \cite{Kurtzer:2016:singularity} containers, + and developing standards as to what inputs and outputs are provided to each component of the workflow. +Another key goal of modularization is to separate the \emph{operator} from the \emph{method}. +Currently, binding calculations are most often done by a human expert, who makes a variety of decisions along the way (though Schr\"odinger's workflow represents real progress toward automation~\cite{schrodinger_accurate_2015}), making it difficult to separate the importance of human expertise from the merits of the methods employed. +Containerizing and modularization will be key for this, allowing methods to be employed only in a well-defined way which is reproducible. It is this type of science -- coupled with benchmark tests -- which is needed to advance the field. + + +\section{THE FUTURE OF THIS WORK} \label{sec:updates} -This work has so far presented a small set of benchmark systems for binding free energy calculations, and has highlighted some of the ways in which they have already proven their utility. -However, the scope of these sets is still quite limited. -More, increasingly diverse, host-guest systems will help probe the strengths and weaknesses of force fields, and to drive their improvement. -At the other end of the spectrum, we need more complex and challenging benchmark sets for proteins including simple models, like T4 lysozyme as well as candidate drug targets. -And there may be community interest in test systems specifically selected to challenge sampling algorithms, without reference to experimental data. - -Several candidate hosts and proteins are worth mentioning in this regard. -Among host-guest systems, there is a particularly extensive experimental literature on cyclodextrins \cite{godinez_thermodynamic_1997, rekharsky_complexation_1998}, and they are tractable computationally ~\cite{henriksen_computational_2015, wickstrom_parameterization_2016}. -As to artificial protein binding sites, the two variants of the CCP protein model binding site~\cite{fitzgerald_ligand-gated_1996, banba_free_2000, banba_efficient_2000, rocklin_blind_2013, musah_artificial_2002, rosenfeld_excision_2002} offer a modest increase in difficulty relative to the T4 lysozyme sites discussed above. -And thrombin and the bromodomains appear to be promising examples of candidate drug targets for inclusion in a growing set of benchmark systems. -Thrombin is a serine protease that has received prior attention from free energy studies~\cite{wang_achieving_2012, schrodinger_accurate_2015, calabro_elucidation_2016}. -Experimental data exhibits interesting trends~\cite{baum_non-additivity_2010} that can partly be explained by simulations~\cite{calabro_elucidation_2016}; but challenges remain~\cite{calabro_accelerating_2015}. -Bromodomains may also be interesting, especially given that relatively high accuracies have been reported, relative to experiment. -At the same time, binding modes may be non-obvious and the diversity of ligands could pose problems for relative free energy calculations~\cite{aldeghi_accurate_2016}. -Other systems will undoubtedly emerge as promising benchmarks as well, and we seek community input to help identify these. - -In order to provide for updates of this material as new benchmark systems are defined, and to enable community input into the process of choosing them, we will make the LaTeX source for this article on GitHub at \url{http://www.github.com/mobleylab/benchmarksets}. -We encourage use of the issue tracker for discussion, comments, and proposed updates. We plan to incorporate new material via GitHub as one would for a coding project, then make it available via a preprint server, likely bioRxiv. +This work has so far presented a small group of benchmarks for binding free energy calculations, highlighted some of the ways in which they have already proven their utility, and has suggested some potential future benchmark systems. +We hope that the community will become involved in identifying, characterizing, and helping to select additional benchmark systems, both from those proposed here as well as from systems which are currently being studied or which we have overlooked. +We seek community input to help characterize, identify and share such systems. +We also expect that there may be community interest in test systems specifically selected to challenge sampling algorithms, without reference to experimental data. + +In order to provide for updates of this material as new benchmark systems are defined, and to enable community input into the process of choosing them, we have made the LaTeX source for this article on GitHub at \url{http://www.github.com/mobleylab/benchmarksets}, with each version having a permanent DOI assigned by Zenodo. +We encourage use of the GitHub issue tracker for discussion, comments, and proposed updates. +We plan to incorporate new material via GitHub as one would for a coding project, then make it available via a preprint server, likely bioRxiv. Given substantial changes to this initial version of the paper, it may ultimately be appropriate to make it available as a ``perpetual review''~\cite{mobley_proposal_2015} via another forum allowing versioned updates of publications. -\section{CONCLUSIONS AND OUTLOOK} +Ideally, we might also update this work in the future with results from ``standard'' calculations on the benchmark systems discussed here, along with links to code to allow reproduction of those calculations. +\section{CONCLUSIONS AND OUTLOOK} Binding free energy calculations are a promising tool for predicting and understanding molecular interactions and appear to have enough accuracy to provide substantial benefits in a pharmaceutical drug discovery context. However, progress is needed to improve these tools so that they can achieve their potential. To achieve steady progress, and to avoid potentially damaging cycles of enthusiasm and disillusionment, we need to understand and be open and honest about key challenges. -Benchmarks are vital for this, as they allow researchers in the field to rigorously test their methods, arrive at a shared understanding of problems, and measure progress on well-characterized yet challenging systems. +Community adoption of well-chosen benchmark systems is vital for this, as it will allow researchers to rigorously test and compare methods, arrive at a shared understanding of problems, and measure progress on well-characterized yet challenging systems. It is also worth emphasizing the importance of sharing information about apparently well thought-out and even promising methods that do \emph{not} work, rather than sharing only what does appear to work. Identifying and addressing failure cases and problems is critically important to advancing this technology, but failures can be harder to publish, and may even go unpublished, even though they serve a unique role in advancing the field. -We therefore strongly encourage that such results be shared and welcomed by the research community. +We therefore strongly encourage that such results be shared and welcomed by the research community. +Potentially, the GitHub repository connected with this perpetual review paper could serve as a place to deposit input files and summary results of tests on these benchmark systems, with summary information perhaps being included in this work itself or topical sub-reviews within the same repository. -Here, we proposed several benchmark systems for binding free energy calculations. +Here, we have proposed several benchmark systems for binding free energy calculations. These embody a subset of the key challenges facing the field, and we plan to expand the set as consensus emerges. Hopefully, these systems will serve as challenging standard test cases for new methods, force fields, protocols, and workflows. Our desire is that these benchmarks will advance the science and technology of modeling and predicting molecular interactions, and that other researchers in the field will contribute to identifying new benchmark sets and updating the information provided about these informative systems. @@ -712,14 +894,11 @@ \section*{DISCLOSURE STATEMENT} D.L.M. is a member of the Scientific Advisory Board for Schr\"{o}dinger, LLC. M.K.G. is a cofounder and has equity interest in the company VeraChem LLC. \section*{ACKNOWLEDGMENTS} -DLM appreciates financial support from the National Institutes of Health (NIH; 1R01GM108889-01) and the National Science Foundation (NSF; CHE 1352608). MKG thanks the NIH for partial support of this work through grant R01GM061300. The contents of this publication are solely the responsibility of the authors and do not necessarily represent the official views of the NIH or the NSF. +DLM appreciates financial support from the National Institutes of Health (NIH; 1R01GM108889-01) and the National Science Foundation (NSF; CHE 1352608). MKG thanks the NIH for partial support of this work through grants R01GM061300, R01GM070064 and U01GM111528. The contents of this publication are solely the responsibility of the authors and do not necessarily represent the official views of the NIH or the NSF. We also appreciate helpful discussions with a huge number of people in the field, including a wide variety of participants at recent meetings such as the 2016 Workshop on Free Energy Methods in Drug Discovery. Conversations with John Chodera (MSKCC), Chris Oostenbrink (BOKU), Julien Michel (Edinburgh), Robert Abel (Schr\"{o}dinger), Bruce Gibb (Tulane), Matt Sullivan (Tulane), and Lyle Isaacs (Maryland) were particularly helpful. -%Remaining todos -%\listoftodos - \bibliographystyle{abbrv} \bibliography{benchmarkset}