shen-lab / bal

Bayesian Active Learning for Optimization and Uncertainty Quantification with Applications to Protein Docking

BAL

Bayesian Active Learning for Optimization and Uncertainty Quantification with Applications in Protein Docking https://pubs.acs.org/doi/abs/10.1021/acs.jctc.0c00476

Versions in branch:

• master: The standard BAL
• noclashv1: The version whose output structure will gurantee \delta vdw<0.

Dependencies:

• C++ 4.8.5 or higher
• cNMA: Download and install the cNMA from "https://github.com/Shen-Lab/cNMA".
• Energy model: Please download the random forest energy model from https://drive.google.com/open?id=17ByuNoYy0t1R8EjuTK_cMyul5K004MHa
• CHARMM: Download the executable 'CHARMM36a1.exe', GBSW: 'radius_gbsw.str', CHARMM27 topology and parameter files and put them into 'dependencies/'.
• Linux Environment

Change directory path

In src/configuration.h, please change the macros as follow:

• src_dir: change to your current "src" path.
• cnma_path: change to your cNMA path.
• scoring_path: change to your 'random_forest.sav' path.
• output_path: change to the directory where you want to place the 'Result' folder which contains all output files.
• charmm_exe, charmm_para, charmm_top, charmm_setup, charmm_gbsw, pdb_crd, crd_pdb: change paths of these macros to their corresponding paths in 'dependencies/'.

Add unbound proteins:

In src/configuration.h, please change the macro 'protein_name' to your 4-letter/digit protein code. Also please change the macro 'protein_path" to the path where the unbound protein is.

If your unbound protein is not in Protein Docking Benchmark 4.0 and you want to get the UQ results, please append its Kd value into 'src/kd_zero' and append the protein name into 'src/kd_list'. Otherwise, you will only get the refined structures and the area under the posterior.

Compile and Run

• Go to 'BAL/src/'.
• Type './complie' to compile.
• Type './for_train' to run BAL.

Output file:

• a.'end'(dir): containing the refined structures: receptor.pdb ligand.pdb

• b.'scores': The final energy (unit in Kcal/mol) of refined structures.

• c.'lgAUC': The log of area under the posterior.

• d.'cond_prob.dat': The conditional probability of refined structure: P(RMSD(x^,x*)<4 | x* \in Mi)

• e. 'Rmsd_dis': The RMSD(x^,x*) distribution of the posterior

• f. 'UQ': The [lb,ub] values

Citation:

@article{cao2020bayesian,
  title={Bayesian active learning for optimization and uncertainty quantification in protein docking},
  author={Cao, Yue and Shen, Yang},
  journal={Journal of chemical theory and computation},
  volume={16},
  number={8},
  pages={5334--5347},
  year={2020},
  publisher={ACS Publications}
}

Contact:

Yang Shen: [email protected]