RF-Score-VS is a novel Random Forest-based scoring function for Virtual Screening which predicts binding affinity. Its descriptors are based on RF-Score developed by Pedro Ballester et. al. Presented binary implements RF-Score-VS v2, meaning, it counts atoms of certain types within a 12A radius, divided into 2A bins. Further information about reported performance in various scenarios and validation across datasets, see the publication.

• Linux 64-bit
• Windows 64-bit
• MacOS 64-bit

• SDF/MDL (.sdf, .mol)
• Mol2 (.mol2)
• PDBQT (.pdbqt)
• PDB (.pdb)

Download package appropriate for your platform, which contains the binary and sample data to test the RF-Score-VS. To use the scoring function uncompress the archive and open a terminal in the same directory as the binary.

• untagged parameters are treated as docked ligands; user can supply multiple molecular files [required]
• -i input file format; if not present then based on extension [optional]
• --receptor a protein file; format based on extension [required]
• -O output file; if -o is not present file format is based on extension [optional]
• -o output file format; if -O is not present then molecules are printed to standard output [optional]

RF-Score-VS predicitons are in -pK units, which means the higher the score the better. To select best binder sort in descending order.

To rescore docked conformations simply run (on Windows omit the leading ./):

./rf-score-vs --receptor protein.pdb ligands.sdf -O ligands_rescored.sdf

Producing CSV files, the RF-Score-VS score is appended to "RFScoreVS_v2" column (additional --field parameter to limit output columns):

./rf-score-vs --receptor protein.pdb ligands.sdf -o csv --field "name" --field "RFScoreVS_v2"

Running test data included with bundle:

./rf-score-vs --receptor test/receptor_rdkit.pdb test/actives_docked.sdf -ocsv

To get the list of all available parameters:

./rf-score-vs --help

NOTE: There is an expected overhead at the beginning of execution of RF-Score-VS binary due to setting up of temporary Python environment.

• Single binary is obtained by freezing the Python script with Pyinstaller
• Open Drug Discovery Toolkit with OpenBabel
• Random forest trees are compiled using sklearn-compiledtrees for extra performance

• Python environment and all dependencies are prepared using Miniconda with Intel MKL optimizations (see the Miniconda LICENSE)
• All components provided here are covered by on permissive 3-clause BSD License.

• Wójcikowski M, Ballester PJ, Siedlecki P. Performance of machine-learning scoring functions in structure-based virtual screening. Sci Rep. Nature Publishing Group; 2017;7: 46710. doi:10.1038/srep46710

• Wójcikowski M, Zielenkiewicz P, Siedlecki P. Open Drug Discovery Toolkit (ODDT): a new open-source player in the drug discovery field. J Cheminform. 2015;7: 5317. doi:10.1186/s13321-015-0078-2

• Ballester PJ, Mitchell JBO. A machine learning approach to predicting protein-ligand binding affinity with applications to molecular docking. Bioinformatics. 2010;26: 1169–1175. doi:10.1093/bioinformatics/btq112

• Ballester PJ, Schreyer A, Blundell TL. Does a more precise chemical description of protein-ligand complexes lead to more accurate prediction of binding affinity? J Chem Inf Model. 2014;54: 944–955. doi:10.1021/ci500091r

• Li H, Leung K-S, Wong M-H, Ballester PJ. Improving AutoDock Vina Using Random Forest: The Growing Accuracy of Binding Affinity Prediction by the Effective Exploitation of Larger Data Sets. Mol Inform. WILEY-VCH Verlag; 2015;34: 115–126. doi:10.1002/minf.201400132