Zoey's way of retrodocking: Difference between revisions
No edit summary |
(Added analysis script.) |
||
| Line 76: | Line 76: | ||
== Retrodocking - analysis == | == Retrodocking - analysis == | ||
'''Connect to the gimel headnode.''' This is currently the only script that requires running on the gimel headnode. Then, run the following: | |||
sh /nfs/home/zdingman/scripts/cluster2_docking/get_aucs_from_xtract.sh [path/to/folder/with/actives.names+decoys.names] | |||
[[Category:Zoey's Way]] | [[Category:Zoey's Way]] | ||
Latest revision as of 22:10, 15 October 2026
The instructions detailed here are my adaptation of the Read-The-Dock-Docs standard procedures, which can be found here. The Docs don't fully cover preparation on cluster2 (i.e., gimel), and this wiki page serves to keep track of my own scripts and post-processing files. Also meant to be a quick-and-dirty list of commands for copy/paste-ing.
Directory preparation
We'll make a directory for your protein structure, with subdirectories for each individual step.
mkdir [protein_name] cd [protein_name] mkdir preparation mkdir blaster mkdir retrodocking
Initial protein preparation
Use Maestro to prepare the protein. Cap termini, protonate, minimize hydrogens (but not heavy atoms!), etc. Save as a pdb file: rec_and_xtal_maestro.pdb
Split your protein using ChimeraX. Separate out the xtal ligand from the rest of the structure. Save as rec_maestro.pdb and xtal_maestro.pdb. Upload these files to the previously created preparation directory on cluster2.
Prepare your protein using prepare_protein.py. See the documentation here. My own fork is currently identical to Andrii's script, but I intend to clean up some of the pathways eventually. Run the following:
python /nfs/home/zdingman/scripts/dockopt/prepare_protein.py --rec rec_maestro.pdb --xtal xtal_maestro.pdb
Blastermaster
Change your directory. Assuming you're still in your preparation folder:
cp rec.pdb ../blaster ; cp working/rec.crg.pdb ../blaster ; cp xtal-lig.pdb ../blaster cd blaster
Source the environment and initialize the job. This is verbatim from Read-The-Dock-Docs:
source /mnt/nfs/soft/dock/versions/dock385/env.sh pydock3 blastermaster - new cd blastermaster_job
Modify blastermaster_config.yaml. Make any edits necessary to your blastermaster_config.yaml file now. Typically we adjust the bump_max from 10.0 to 100.0.
Run the job. Run the following:
pydock3 blastermaster - run
Retrodocking - job launching
Below are the steps for running any sort of dock job on cluster2 -- in this case, a retrodocking job. However, the procedure is identical for any larger job you may wish to submit.
Change to your docking directory. Assuming you're still in the blastermaster folder. Run the following:
cd ../../retrodocking
Prepare your directory. Make the requisite directories. Add your own sdi file (formatted where each line is the full path to a tgz package of built molecules) to the sdi folder, and call it mols.sdi.
mkdir output ; mkdir sdi cp -r ../blaster/blastermaster/dockfiles/ . cp /nfs/home/zdingman/scripts/cluster2_docking/super_gimel_dock38.sh .
Launch the jobs. Run the following:
sh super_gimel_dock38.sh sdi/mols.sdi output/
Retrodocking - extraction
Confirm all jobs have completed. Run squeue -u [your_username] to check if you have any outstanding jobs. Check the integrity of your OUTDOCK files, and the contents of your `.err` files as well. A simple wc -l output/*/*err usually suffices.
Generate your dirlist. Run the following:
source /nfs/soft/dock/versions/dock37/DOCK-3.7.5.0/env.sh ; source /nfs/soft/python/envs/complete/current/env.sh sh /mnt/nfs/exa/work/ak87/PROGRAM/LSD/make_dirlist.sh sh /mnt/nfs/exa/work/ak87/PROGRAM/LSD/cat_getposes.sh dirlist
Extract all outdock files and their corresponding poses. The below script retrieves everything in the dirlist which has a dock score of +100 or less:
source /mnt/nfs/home/ak87/miniconda3/bin/activate /mnt/nfs/home/ak87/miniconda3/envs/py2 python $DOCKBASE/analysis/extract_all_blazing_fast.py dirlist extract_all.txt 100 python $DOCKBASE/analysis/getposes_blazing_faster.py '' extract_all.sort.uniq.txt 2000 output_name.mol2 test.mol2.gz
Retrodocking - analysis
Connect to the gimel headnode. This is currently the only script that requires running on the gimel headnode. Then, run the following:
sh /nfs/home/zdingman/scripts/cluster2_docking/get_aucs_from_xtract.sh [path/to/folder/with/actives.names+decoys.names]