AlloPred

This is the code and training/testing data used at the AlloPred web server, which predicts allosteric pockets on proteins from a PDB format file:

http://www.sbg.bio.ic.ac.uk/allopred/home

The server is currently having problems due to configuration issues. We are looking to resolve this.

The method is described here: Greener, JG and Sternberg, MJE, AlloPred: prediction of allosteric pockets on proteins using normal mode perturbation analysis, BMC Bioinformatics, 2015, 16(335). Link to paper.

Downloading this code lets you run AlloPred locally. It should work on all Unix-based systems. With modification it could be made to work on Windows.

Please contact Joe Greener (http://jgreener64.github.io) for support, queries and suggestions. Alternatively, open an issue on GitHub.

Requirements

• Python 2.7 with the NumPy and ProDy packages installed.
• fpocket v2.0, which can be downloaded from here. Follow the installation instructions to compile the executables.
• SVM-light, which can be downloaded from here. Follow the installation instructions to compile the executables.

Usage

Follow these steps to set up AlloPred - the shell commands are for bash:

1. Download the files and extract them as usual, or clone the repository.

2. The environmental variables $ALLOPRED_DIR and $SVM_LIGHT_DIR need to be set as the filepaths to the AlloPred directory and the SVM-light directory respectively:

export ALLOPRED_DIR=/path/to/allopred/
export SVM_LIGHT_DIR=/path/to/svm_light/

Consider adding these lines to your profile so you don't have to run them every session.

Follow these step to run AlloPred:

1. Obtain a PDB format file (in_file.pdb), e.g. from the Protein Data Bank.

2. Create a one-line file (act_res.txt) containing the active site residues of the protein. The format is 10:A,11:B for residue 10 on chain A and residue 11 on chain B. These can be found using resources such as the Catalytic Site Atlas. An example PDB file and active residue file can be found in the example directory of AlloPred.

3. Run fpocket v2.0 on the PDB file:

fpocket -f in_file.pdb

This assumes fpocket is on the path. This produces the directory in_file_out. AlloPred is optimised on the default fpocket parameters but you can change these in accordance with the fpocket documentation if you wish.

4. The following command, from the directory containing in_file.pdb and in_file_out, runs the AlloPred pipeline:

python $ALLOPRED_DIR/run_allopred.py in_file act_res.txt

The arguments are the input file prefix and the path to the active site residue file. Running the run_allopred.py script with fewer than 2 arguments returns these instructions for the command.

5. The output files are:

• in_file.out: the AlloPred output file containing the input parameters and the values for each pocket in order of AlloPred ranking.
• in_file.svm: the SVM input file in the SVM-light format.

Other files

• dataset contains information on the training and testing sets.
• example contains the inputs and outputs of an example run using the PDB entry with ID 1FX2.
• svm_model.txt is the optimised SVM built on the whole training set.

Reproducibility

For reference, here are steps to reproduce the results in Figure 3 of the paper (link above):

1. Obtain PDB files from the test set list in dataset/test_set.tsv.
2. Keep only the chains for each file given in dataset/test_set.tsv.
3. For each protein carry out steps 2-4 above using the active site residues given in dataset/test_set.tsv.
4. Compare the residues for the top predicted pockets in in_file.out to the allosteric residues given in dataset/test_set.tsv.

Note

On some systems the normal mode calculation step fails (More than 6 zero eigenvalues are calculated). This appears to be due to an error in scipy.linalg.eigh which causes an error in the calcModes function of ProDy.

Thank you for using AlloPred!