scoring

Scoring Decoys

This directory contains all the scripts needed to calculate the GDT-TS scores of a decoy dataset. The process of calculating the scores is as follow:

1. Create a new dataset folder in the /data/ directory, and copy your decoy PDBs into a folder called 'pdbs' within your new dataset folder. Copy the target crystal structure used for the comparison and make sure its named the same as the newly created dataset folder plus the extension '.pdb'. For more information on how this should look like, please refer to the README within the /data/ directory.

2. Open calculate_GDTTS.py and change the data_folder variable to the path to the dataset directory. The path should be set relative to this file. This script utilizes the LGA package which is the standard tool used to calculate GDT-TS. This process stores the output results from the LGA package into a folder named '/scores/' within your dataset folder. Scoring is parallelized with MPI. The score.csh batch script can be used to submit the job to Cori. Make sure to change number of nodes, number of cores, and total time according to your dataset. This process takes one hour to run using 128 Haswell cores on Cori for a dataset of size 300,000.

IMPORTANT : You may need to change the lga_command variable within the script if the decoy structures do not align with target crystal structure. If they do not align, remove the -sda flag from the command and insert the flags -aa1:n1:n2 for the decoy structure and -aa2:n1:n2 for the target structure, replacing n1 and n2 with the residue index range for each given structure.

3. Open calculate_GDTTS.py and change the data_folder variable to the path to the dataset directory. This script will read all the LGA output files generated from the last script and create a CSV file containing the PDB ids and corresponding GDT-TS scores saved in the dataset directory.