This is an example of generating ROC curves from virtual screening data generated using rDock.
It's based on the example provided by rDock that can be found here.
It is not a complete example as it uses already generated docking scores that are contained in the file hivpr_all_results.sd.gz
Docker must be installed on the host machine.
The hivpr_all_results.sd.gz input file is too large to store in GitHub so we download it instead. We also grab ligands.txt and decoys.txt that define which ligands are actives and which are decoys.
1_download_data.shThe result is the downloaded hivpr_all_results.sd.gz, ligands.txt and decoys.txt files.
This filters the docked structures to find the best score for each structure and then generates the input that is needed by R.
./2_prepare_data.shThis takes a few minutes The result is the file hivpr_1poseperlig.sdf.gz which contains the best pose for each ligand and data extracted from that file that is needed by R (the file dataforR_uq.txt).
This uses the informaticsmatters/rdock-mini Docker image that contains the rDock programs.
This uses the data from the previous step to geneate a JPG with the ROC curve.
./3_generate_roc.shThe result is the JPG file hivpr_Rinter_ROC.jpg
This uses the informaticsmatters/r-roc Docker image that contains R and the ROCR package.
- Create a more complete example based on data from DUD-E.
- Look into generating outputs in different formats.
- Look into generating ROC curves that combine/compare different datasets.