search

rdk / p2rank

P2Rank: Protein-ligand binding site prediction from protein structure based on machine learning.
MIT License
251 stars 34 forks source link
binding-sites bioinformatics chimerax computational-biology docking drug-discovery groovy java machine-learning mmcif molecular-structures p2rank pdb protein-ligand-docking protein-ligand-interactions protein-structure proteins pymol structural-bioinformatics virtual-screening

readme

P2Rank

Ligand-binding site prediction based on machine learning.

P2Rank illustration

Build Status License: MIT GitHub all releases

Description

P2Rank is a stand-alone command-line program for fast and accurate prediction of ligand-binding sites from protein structures. It achieves high prediction success rates without relying on external software for computation of complex features or on a database of known protein-ligand templates.

✨ What's new?

Requirements

P2Rank is tested on Linux, macOS, and Windows.

Setup

P2Rank requires no installation. Binary packages are available as GitHub Releases.

Usage

prank predict -f test_data/1fbl.pdb         # predict pockets on a single pdb file

See more usage examples below...

Algorithm

P2Rank makes predictions by scoring and clustering points on the protein's solvent accessible surface. Ligandability score of individual points is determined by a machine learning model trained on a dataset of known protein-ligand complexes. For more details, see the slides and publications.

Presentation slides introducing the original version of the algorithm: Slides (pdf)

Publications

If you use P2Rank, please cite relevant papers:

👨‍💻 Usage Examples

Following commands can be executed in the installation directory.

Help and version

prank help   # print help for main commands and parameters
prank -v     # print version and some system info

Predict ligand binding sites (P2Rank algorithm)

prank predict test.ds                    # run on dataset containing a list of pdb/cif files

prank predict -f test_data/1fbl.pdb      # run on a single pdb file
prank predict -f test_data/1fbl.cif      # run on a single mmCIF file
prank predict -f test_data/1fbl.bcif     # run on a single BinaryCIF file
prank predict -f test_data/1fbl.pdb.gz   # run on a single gzipped pdb file (other formats can be compressed too)
prank predict -f test_data/1fbl.cif.zst  # run on a single cif file compressed with Zstandard 

prank predict -threads 8     test.ds     # specify num. of working threads for parallel dataset processing
prank predict -o output_here test.ds     # explicitly specify output directory

prank predict -c alphafold   test.ds     # use alphafold config and model (config/alphafold.groovy)  
                                         # this profile is recommended for AlphaFold models, NMR and cryo-EM 
                                         # structures since it doesn't depend on b-factor as a feature

Prediction output

For each structure file {struct_file} in the dataset, P2Rank generates several output files:

Configuration

You can override the default parameter values in a custom config file:

prank predict -c config/example.groovy  test.ds
prank predict -c example                test.ds # same effect, config/ is default location and .groovy implicit extension

It is also possible to override parameters on the command line using their full name after - (not --).

prank predict                   -visualizations 0 -threads 8  test.ds   #  turn off visualizations and set the number of threads
prank predict -c example.groovy -visualizations 0 -threads 8  test.ds   #  overrides defaults as well as values from example.groovy

P2Rank has many configurable parameters. To see the list of standard parameters look into config/default.groovy and other example config files in this directory. To see the complete commented list of all (including undocumented) parameters see Params.groovy in the source code.

Rescoring (PRANK algorithm)

In addition to predicting new ligand binding sites, P2Rank is also able to rescore pockets predicted by other methods (Fpocket, ConCavity, SiteHound, MetaPocket2, LISE, DeepSite, and PUResNetV2.0 are supported at the moment).

Rescoring output:

prank rescore fpocket.ds                   
prank rescore fpocket.ds -o output_here   # explicitly specify output directory
prank rescore fpocket.ds -c rescore_2024  # use new experimental rescoring model (recommended for alphafold models)

For rescoring, the dataset file needs to have a specific 2-column format. See examples in test_data/: fpocket.ds, concavity.ds, puresnet.ds.

New experimental rescoring model -c rescore_2024 shows promising result but hasn't been fully evaluated yet. It is recommended for AlphaFold models, NMR and cryo-EM structures since it doesn't depend on b-factor as a feature.

Run fpocket and rescore in one command

You can use fpocket-rescore command to run Fpocket and then rescore its predictions automatically.

prank fpocket-rescore test.ds                                       # expects 'fpocket' command in PATH
prank fpocket-rescore test.ds -fpocket_command "/bin/fpocket -w m"  # specify custom fpocket command (optionally with arguments)
prank fpocket-rescore test.ds -fpocket_keep_output 0                # delete fpocket output files

In this case, the dataset file can be a simple list of pdb/cif files since Fpocket predictions will be calculated ad-hoc. prank fpocket-rescore will produce predictions.csv as well, so it can be used as an in-place replacement for prank predict in most scenarios. Note: if you use fpocket-rescore, please cite Fpocket as well.

Evaluate prediction and rescoring models

Use following commands to calculate prediction metrics (prediction success rates using DCA, DCC, ...) on structure files, where the ligands are present.

prank eval-predict -f test_data/1fbl.pdb         # evaluate default prediction model on a single file
prank eval-predict              test.ds          # evaluate default prediction model on a dataset with known ligands
prank eval-predict -c alphafold test.ds          # evaluate specific prediction model on a dataset with known ligands

prank eval-rescore                 fpocket.ds    # evaluate default rescoring model on a dataset with known ligands
prank eval-rescore -c rescore_2024 fpocket.ds    # evaluate specific rescoring model on a dataset with known ligands

🏗️ Build from sources

This project uses Gradle build system via included Gradle wrapper. On Windows, use bash to run build commands (installed by default with Git for Windows).

git clone https://github.com/rdk/p2rank.git && cd p2rank
./make.sh       

./unit-tests.sh    # optionally you can run tests to check everything works fine on your machine        
./tests.sh quick   # runs further tests

Now you can run the program via:

distro/prank       # standard mode that is run in production
./prank.sh         # development/training mode

To use ./prank.sh (development/training mode) first you need to copy and edit misc/locval-env.sh into repo root directory (see training tutorial).

⚖️ Comparison with Fpocket

Fpocket is a widely used open source ligand binding site prediction program. It is fast, easy to use and well documented. As such, it served as a great inspiration for this project.

Some practical differences:

Both Fpocket and P2Rank have many configurable parameters that influence behaviour of the algorithm and can be tweaked to achieve better results for particular requirements.

🙌 Thanks

This program builds upon software written by other people, either through library dependencies or through code included in its source tree (where no library builds were available). Notably:

🤝 Contributing

We welcome any bug reports, enhancement requests, and other contributions. To submit a bug report or enhancement request, please use the GitHub issues tracker. For more substantial contributions, please fork this repo, push your changes to your fork, and submit a pull request with a good commit message.