Use your custom global optimization algorithm to predict the three-dimensional structure of a peptide from its amino acid sequence. Using multivariate quantile function the continuous search space formed by Rosetta fragments and Dunbrack rotamer library. The search space is unit hypercube.
Results using Adaptive Differential Evolution. Native (green) vs predicted (cyan).
To compile from source, you will need:
- Some Linux distro (Ubuntu/Arch Linux/etc).
- C++ compiler with C++17 support (gcc/clang).
- The Rosetta software suite.
- Header-only library mveqf.
- OpenMP for multithreading.
- Get the Rosetta weekly source package and build it with SCons.
- Download the source code:
git clone https://github.com/poluyan/PEPSGO
cd PEPSGO/
- Modify the CMakeLists.txt. Set the path to the Rosetta build with the kernel and gcc versions which were used in Rosetta build:
set( ROSETTAMAINPATH "/work/rosetta_src_2020.50.61505_bundle/main" )
set( ROSETTALINUXVER "5.9" )
set( ROSETTACPPCOMP "gcc" )
set( ROSETTACPPVER "10.2" )
set( MVEQFDIR "${CMAKE_SOURCE_DIR}" )
- Then compile PEPSGO from source with
cmake .andmakecommands. This will generate library filelibpepsgo.soand demos presented inbindirectory. - The file
run_testX.shin each test process the followinginputdirectory and get the proper input for the program.
sequence.fastaPeptide sequence in FASTA format.fragments.NmersFragment file formatted in Rosetta-style (like the output from fragment picker). Fragments can be any length.native.dataFile with a path to the native structure.prediction.horizSecondary structure prediction in PSIPRED HFORMAT.
Only peptide sequence in FASTA format is requred for most examples presented in demo.
The demos/testX contains examples of using PEPSGO.
The PEPSGO is distributed under Apache License 2.0 and it is open-source software. Feel free to make a copy and modify the source code, but keep the copyright notice and license intact. The PEPSGO is Rosetta-based, please read Rosetta Licensing Information before using.