Prediction of protein density using Random Forest Regressors. Released under GNU General Public License v3.0.
Contains classes for sequence and structure which allow for the easy estimation of protein density from sequences (either from PDB files, FASTA files, three-letter code sequence strings or one-letter code sequence strings) or structure-derived charactieristics (from PDB files). Automatically calculates the relevant features for the protein structure or sequence at a selected temperature (300K or 310.15K)
Provides an easy-to-use interface for the use of DensiTree.py designed for use with Google Colab, which allows for file uploads and text input of sequences. All dependencies are installed to create a conda environment consistent with itself and the trained random forest regressors to allow for easy prediction of protein density.
(From random_forest/info.txt)
A range of random forest regressors trained on different features at both 310.15 K and 300 K are supplied here. When temperature is not noted, the temperature is 300 K as standard. The random forests are trained on 1ns simulations of proteins from a protein structure dataset derived from the Protein-Protein Docking Benchmark 5 (see: Vreven, Thom, Iain H. Moal, Anna Vangone, Brian G. Pierce, Panagiotis L. Kastritis, Mieczyslaw Torchala, Raphael Chaleil et al. "Updates to the integrated protein–protein interaction benchmarks: docking benchmark version 5 and affinity benchmark version 2." Journal of molecular biology 427, no. 19 (2015): 3031-3041.) 21 structures captured between 0 and 1 ns inclusive of a production run of an MD simulation for each protein structure are extracted.
- RF_300_all is trained on the mean values of the mean values of all features (see figure S21 for a full list of features) for each protein in the 260-protein dataset.
- RF_300_20 (AKA: struct_RF) is trained on the mean values of the 20 features found to be of the most importance by the ablation study (see figure S19 and S22).
- RF_300_seq is trained on only the sequence data (the percentage composition of the protein sequence by residue) of each protein.
310.15K folder contains results obtained from 310.15 K simulations. The feature values consist of a range of physical characteristic values of a protein, ranging from the Solvent Accessible Surface Area to the percentage of residues in a coil configuration.
- RF_310_all is trained on all extracted features.
- RF_310_20 is trained on the mean values of the 20 features found to be of the most importance by the ablation study (see figure S19 and S22).
- RF_310_seq is trained on only the sequence data (the percentage composition of the protein sequence by residue) of each protein.