P2Rank returns very many pocket predictions (AlphaFold mode)

[gabedol]

• Hi, I'm running p2rank of AFDB. One of the things i'm noticing is just how many pockets it returns for some proteins. For example, when I run it on AF-Q8TEX9, I get 14 pockets (written below). Is this to be expected? What kinds of score filtering do you employ? Are p2rank predictions on AFDB equally well-calibrated/accurate? Here is the command I ran:

p2rank_2.5.1/prank predict -c p2rank_2.5.1/config/alphafold.groovy -f "$pdb" -o AF2_prank_predictions

name , rank, score, probability, sas_points, surf_atoms, center_x, center_y, center_z, residue_ids, surf_atom_ids

pocket1 , 1, 10.40, 0.482, 67, 37, -1.5809, 8.2592, -15.3101, A_311 A_312 A_325 A_331 A_334 A_335 A_338 A_370 A_374 A_378 A_412 A_413 A_416 A_420 A_457, 2436 2437 2439 2447 2553 2554 2556 2597 2598 2600 2603 2623 2625 2626 2632 2634 2657 2915 2940 2941 2967 3224 3225 3226 3230 3231 3233 3251 3253 3254 3255 3256 3257 3258 3282 3580 3581

...

pocket13 , 13, 1.57, 0.020, 25, 10, -4.2693, -12.5112, -7.6807, A_546 A_584 A_585 A_588 A_675, 4269 4270 4271 4272 4555 4561 4584 4586 5261 5262
pocket14 , 14, 1.08, 0.007, 18, 18, 40.9840, 6.8970, -20.0550, A_133 A_136 A_137 A_140 A_91 A_92 A_94 A_97, 724 731 733 746 748 751 771 773 1056 1082 1084 1085 1086 1087 1088 1089 1091 1115

[rdk]

Hi @gabedol, thanks for the interest. Please read #76 and let me know if you have any remaining questions.

[gabedol]

Thanks, this makes sense. One remaining question – when running on a PDB file with ligands, do you recommend removing ligands to find the corresponding pocket?

[rdk]

In the typical case no, you can leave the file as is. P2Rank builds the prediction surface from amino acid residues only, so any HETATM ligands present in the input are ignored. (They can be used as ground truth, but only in eval mode.)

One advanced caveat: covalently bound ligands and modified residues are a more nuanced case, see #79 for that discussion.

[gabedol]

Thank you, I just have one more issue – when running, I get this error:
[INFO] FeatureSetup - enabledFeatures: [chem, volsite, protrusion, bfactor, atom_table] [0/510]
WARNING: A terminally deprecated method in sun.misc.Unsafe has been called
WARNING: sun.misc.Unsafe::objectFieldOffset has been called by jsr166y.ForkJoinPool (file:/Genomics/pritykinlab/gabe/jupys/chemprot/pocket-prediction/p2rank_2.5.1/bin/lib/jsr166y-1.7.0.jar)
WARNING: Please consider reporting this to the maintainers of class jsr166y.ForkJoinPool
WARNING: sun.misc.Unsafe::objectFieldOffset will be removed in a future release
Can it be ignored?

[rdk]

Yes, this can be ignored.

Could you share the exact versions of P2Rank, OS and JVM you are using so I can fix the warning?