Dockopt (pydock3 script)

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

The name "dockopt", aside from being an uncreative rehash of the name "blastermaster", derives from the notion of a literal dockopt, i.e., the person in charge of a dock who manages freight logistics and bosses around numerous dockworkers. In this analogy, a single dockworker corresponds to the processing of a single docking configuration.

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>

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 the short cache and long cache should be located. E.g.:

export SHRTCACHE=/dev/shm  # temporary storage for job files
export LONGCACHE=/dev/shm  # long-term storage for files shared between jobs

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, 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>

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.

You can also set the following flags to adjust retrodock job submission behavior. This example show the default values:

pydock3 dockopt - run <JOB_SCHEDULER_NAME> --retrodock_job_max_reattempts=0 --retrodock_job_timeout_minutes=None

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 each retrodock job directory, there are the following files and sub-directories:

• working/: intermediate files
• dockfiles/: parameters files and INDOCK for given docking configuration
• output/: contains:
  • joblist
  • sub-directories 1/ for actives and 2/ for decoys (each containing OUTDOCK and test.mol2 files)
  • log files for the retrodock jobs
• retrodock_job_results.csv: data loaded from OUTDOCK files for both actives and decoys
• roc.png: the ROC enrichment curve (log-scaled x-axis) for given docking configuration