Dockopt (pydock3 script)

From DISI
Revision as of 07:17, 28 October 2022 by Ianscottknight (talk | contribs)
Jump to navigation Jump to search

dockopt allows the generation of many different docking configurations which are then evaluated & analyzed in parallel using a specified job scheduler (e.g. Slurm).

Note for UCSF Shoichet Lab members

pydock3 is already installed on the following clusters. You can source the provided Python environment scripts to expose the pydock3 executable:

Wynton

 source /wynton/group/bks/soft/python_envs/python3.8.5.sh

Gimel

Only nodes other than gimel itself are supported, e.g., gimel5.

ssh gimel5
source /nfs/soft/ian/python3.8.5.sh

init

First you need to create the file structure for your dockopt job. To do so, simply type

pydock3 dockopt - init

By default, the job directory is named dockopt_job. To specify a different name, type

pydock3 dockopt - init [--job_dir_name="dockopt_job"]

The job directory contains two sub-directories:

  1. working: input files, intermediate blaster files, sub-directories for individual blastermaster subroutines
  2. retrodock_jobs: individual retrodock jobs for each docking configuration

The key difference between the working directories of blastermaster and dockopt is that the working directory of dockopt may contain multiple variants of the blaster files (prefixed by a number, e.g. "1_box"). These variant files are used to create the different docking configurations specified by the multi-valued entries of dockopt_config.yaml. They are created efficiently, such that the same variant used in multiple docking configurations is not created more than once.

If your current working directory contains any of the following files, then they will be automatically copied into the working directory within the created job directory. This feature is intended to simplify the process of configuring the dockopt job.

  • rec.pdb
  • xtal-lig.pdb
  • rec.crg.pdb
  • reduce_wwPDB_het_dict.txt
  • filt.params
  • radii
  • amb.crg.oxt
  • vdw.siz
  • delphi.def
  • vdw.parms.amb.mindock
  • prot.table.ambcrg.ambH

Only the following are required. Default versions / generated versions of the others will be used instead if they are not detected.

  • rec.pdb
  • xtal-lig.pdb

If you would like to use files not present in your current working directory, copy them into your job's working directory, e.g.:

cp <FILE_PATH> <JOB_DIR_NAME>/working/

Finally, configure the dockopt_config.yaml file in the job directory to your specifications. The parameters in this file govern the behavior of dockopt.

Note: The dockopt_config.yaml file differs from the blastermaster_config.yaml file in that every parameter of the former may accept either a single value or a list of comma-separated values, which indicates a pool of values to attempt for that parameter. Multiple such multi-valued parameters may be provided, and all unique resultant docking configurations will be attempted.

Single-valued YAML line format:

distance_to_surface: 1.0

Multi-valued YAML line format:

distance_to_surface: [1.0, 1.1, 1.2, 1.3, 1.4, 1.5, 1.6, 1.7, 1.8, 1.9]

Environmental variables

Designate where temporary job files should be placed. E.g.:

export TMPDIR=/scratch

In order for dockopt to know which scheduler it should use, please configure the following environmental variables according to which one of the job schedulers you have.

Slurm

E.g., on the UCSF Shoichet Lab Gimel cluster (on any node other than 'gimel' itself, such as 'gimel5'):

export SBATCH_EXEC=/usr/bin/sbatch
export SQUEUE_EXEC=/usr/bin/squeue

SGE

E.g., on the UCSF Wynton cluster:

export QSTAT_EXEC=/opt/sge/bin/lx-amd64/qstat
export QSUB_EXEC=/opt/sge/bin/lx-amd64/qsub

The following is necessary on the UCSF Wynton cluster:

export SGE_SETTINGS=/opt/sge/wynton/common/settings.sh

On most clusters using SGE, this will probably be:

export SGE_SETTINGS=/opt/sge/default/common/settings.sh

run

Once your job has been configured to your liking, navigate to the the job directory and run dockopt:

cd <JOB_DIR_NAME>
pydock3 dockopt - run <JOB_SCHEDULER_NAME> [--retrodock_job_timeout_minutes=None] [--retrodock_job_max_reattempts=0]

where <JOB_SCHEDULER_NAME> is one of:

  • sge
  • slurm

This will execute the many dockopt subroutines in sequence, except for the retrodock jobs run on each docking configuration, which are run in parallel via the scheduler. The state of the program will be printed to standard output as it runs.

Once the dockopt job is complete, the following files will be generated in the job directory:

  • dockopt_job_report.pdf: contains (1) roc.png of best retrodock job, (2) box plots of enrichment for every multi-valued config parameter, and (3) heatmaps of enrichment for every pair of multi-valued config parameters
  • dockopt_job_results.csv: enrichment metrics for each docking configuration

In addition, the best retrodock job will be copied to its own sub-directory best_retrodock_job/.

Within best_retrodock_job, there are the following files and sub-directories:

  • dockfiles/: parameters files and INDOCK for given docking configuration
  • output/: contains:
    • joblist
    • sub-directories 1/ for actives and 2/ for decoys (the former containing OUTDOCK and test.mol2 files, the latter containing just OUTDOCK)
    • log files for the retrodock jobs
  • roc.png: the ROC enrichment curve (log-scaled x-axis) for given docking configuration

Note: by default, a mol2 file is exported only for actives (output/1/), not for decoys (output/2/), in order to prevent disk space issues.