Rescoring with PLOP

From DISI
Jump to: navigation, search

Instruction for rescoring:

On broadway, copy niu's .tchrc to your directory, make sure you are also using tcsh, not csh or bsh.

1. receptor structure preparation

first, run ~nhuang/scripts/nwdock_pdb_fix rec.pdb rec.new compare the original rec.pdb and rec.new to make sure rec.new has chain name. then move rec.new to rec.pdb.

copy ~nhuang/scripts/rec_h_opt.con, delete the "template ..." line if there has no cofactor.

you can prepare the receptor file either by running PLOP in queue or interactively. I recommend the later one.

Check which node is free by run "checknodes", then login the free one (assume 15) by "ssh klient15"

go to the right directory and run "echo rec_h_opt.con | ~nhuang/plop/plop &"

Note, weird thing going on if you run it directly from broadway headnode which not suppose to be, however, for safety reason not kicked out of broadway, DO LOG INTO CLIENT NODE to run PLOP jobs.

2. ligand parameters files generation

On broadway headnodes, if test.eel1.gz (known ligands) and s_.new.eel1.gz (decoys or whatever) in current directory, run "prep_templates_qgen jobname", a file named jobname.que will be generated, take a look what inside; then do "qsub jobname.que".

for 25,000 ligands, it takes ~ one day to generate parameter files and return one file "jobname.tar.gz".

You may want to check what's inside this package, or not.

3. on Baton2

scp the rec_h_opt.pdb, rec_h_opt.log and jobname.tar.gz to Baton2, untar the tar file, and make sure the directories are consistent (check /ccpr1/nhuang/testcase/new/thrombin_opls/, the subdirectory ligands and mddr_decoys)

run "super_submit_2006.csh jobname opls/amsol &"

then all the individual jobs should be submitted to queue automatically. Once it done, we can take a look together.