SUBDOCK DOCK3.8
Important note- although DOCK 3.8 is in the header of this article, SUBDOCK is perfectly capable of running DOCK 3.7 workloads, though some features of DOCK 3.8 will not be taken advantage of.
Installing
git clone https://github.com/docking-org/SUBDOCK.git
IMPORTANT: subdock.bash expects to live in the same directory as rundock.bash!!!
subdock.bash is located @ subdock.bash relative to the repository root.
subdock.bash can be called directly from any location- it is not sensitive to the current working directory.
What's New?
For those of you that have used a subdock utility before, here's what is new in this release:
1. All jobs platforms (e.g slurm, sge) are supported on the same script
2. GNU Parallel is now supported as a jobs platform! Ideal for small-scale local testing. https://www.gnu.org/software/parallel/
3. Subdock can now be run on both db2.gz individual files & db2.tgz packages. A batch_size can be set for both types, allowing for more flexibility.
4. Arguments can be provided environmentally, e.g "export KEY=VALUE" or on the command line e.g "--key=value"
5. Subdock now prints out a superscript to copy-paste on success, convenient for re-submission.
6. Fully restartable on all jobs platforms! See below section for an explanation on what this means, why it matters, and instructions on usage.
Supported Platforms
There are three platforms currently supported:
1. SLURM
2. SGE (Sun Grid Engine)
3. GNU Parallel (for local runs- ideal for testing)
One of these platforms must be specified- SLURM is the default. These platforms can be set by the
--use-slurm=true --use-sge=true --use-parallel=true
Arguments, respectively
Supported File Types
DOCK can be run on individual db2.gz files or db2.tgz tar packages.
The file type can be specified via the --use-db2=true or --use-db2-tgz=true arguments. db2.tgz is the default
Each job dispatched by SUBDOCK will consume BATCH_SIZE files, where BATCH_SIZE is equal to --use-db2-batch-size or --use-db2-tgz-batch-size depending on which file type is chosen.
The number of jobs dispatched by SUBDOCK is equal to ceil(N / BATCH_SIZE), where N is the total number of input files.
Restartability
Restartability means that we can impose arbitrary time limits on how long our jobs can run *without* losing our progress. Time limits can be as large or as small as we want them to be, even as little as a few minutes per job! This flexibility lets docking jobs efficiently fill in the gaps between longer-running jobs on the same ecosystem, thus they will be preferentially treated by whichever system is in charge of scheduling.
How to use for your Job Platform
On SLURM, runtime can be defined with the "--time" argument, e.g:
subdock.bash --use-slurm=true --use-slurm-args="--time=00:30:00"
This will allow our job to run for 30 minutes before progress is saved & copied out.
On GNU parallel this is accomplished with "--timeout", e.g:
subdock.bash --use-parallel=true --use-parallel-args="--timeout 1800"
On SGE, the same can be achieved using the s_rt and h_rt parameters, e.g:
subdock.bash --use-sge=true --use-sge-args="-l s_rt=00:29:30 -l h_rt=00:30:00"
This tells SGE to warn the job 30 seconds prior to the 30 minute hard limit. GNU and SLURM platforms will provide a hard-coded 30 seconds notice, whereas this notice period must be manually defined for SGE jobs.
How to continue jobs
Run subdock.bash again with the same parameters (particularly EXPORT_DEST, INPUT_SOURCE, USE_DB2, USE_DB2_TGZ, USE_DB2_BATCH_SIZE, and USE_DB2_TGZ_BATCH_SIZE) to restart your jobs! If you saved the superscript SUBDOCK spits out on successful submission, you can simply call that.
You'll know there is no more work to be done if SUBDOCK prints "all N jobs complete!", SUBDOCK will also tell you what proportion of jobs have not yet completed on each submission.
Be careful not to overlap your submissions- there are no guardrails in place to prevent this from happening if you are not careful.
Full Example - All Steps
This example assumes you have access to a DOCK executable, but nothing else.
1. Source subdock code from github
git clone https://github.com/docking-org/SUBDOCK.git
2. Fetch dockfiles from DUDE-Z- we will use DRD4 for this example.
# note- SUBDOCK automatically detects your DOCK version & corrects the INDOCK header accordingly wget -r --reject="index.html*" -nH --cut-dirs=2 -l1 --no-parent https://dudez.docking.org/DOCKING_GRIDS_AND_POSES/DRD4/dockfiles/
3a. Get db2 database subset sample via ZINC-22. Example provided below:
wget http://files.docking.org/zinc22/zinc-22l/H17/H17P050/a/H17P050-N-laa.db2.tgz wget http://files.docking.org/zinc22/zinc-22l/H17/H17P050/a/H17P050-N-lab.db2.tgz wget http://files.docking.org/zinc22/zinc-22l/H17/H17P050/a/H17P050-N-lac.db2.tgz
You can select a db2 database subset via cartblanche22.docking.org- for wget-able files, choose the DOCK37 (*.db2.tgz) format, with URL download type. Multiple download types are supported, for example if you are on Wynton you can download Wynton file paths- removing the need to download the files yourself.
3b. If you downloaded the db2.tgz files yourself, create an sdi.in file from your database subset, which will serve as a list of files to evaluate. For example:
find $PWD -type f -name '*.db2.tgz' > sdi.in
4. Export the parameters we just prepared as environment variables. You need a DOCK executable! This can be found via our download server if you have a license, otherwise lab members can directly pull https://github.com/docking-org/dock3.git. On BKS cluster, some curated executables have been prepared with labels @ /nfs/soft/dock/versions/dock38/executables. DOCK 3.7 executables may be found here as well!
export INPUT_SOURCE=$PWD/sdi.in export EXPORT_DEST=$PWD/output export DOCKFILES=$PWD/dockfiles export DOCKEXEC=/nfs/soft/dock/versions/dock38/executables/dock38_nogist
5. Choose a platform. You must select only one platform - mixing and matching is not supported.
export USE_SLURM=true|... export USE_SGE=true|... export USE_PARALLEL=true|...
Any value other than exactly "true" will be interpreted as false.
6a. Run docking!
bash ~/SUBDOCK/subdock.bash
6b. You can also use command line arguments instead of environment export, if desired. These can be mixed and matched.
export DOCKEXEC=$PWD/DOCK/ucsfdock/docking/DOCK/dock64 bash ~/SUBDOCK/subdock.bash --input-source=$PWD/sdi.in --export-dest=$PWD/output --dockfiles=$PWD/dockfiles --use-slurm=true
7. After executing subdock, it will print out a convenient "superscript" to copy & paste, for any future re-submissions.
Error Messages in my OUTDOCK!
If you're running DOCK 3.8 against recently built ligands, you may encounter error messages that look like this:
1 2 bonds with error Error. newlist is not big enough
If these messages bother you use the dock38_nogist executable described in How_to_install_DOCK_3.8#Prebuilt_Executable
This version voids the code related to the GIST scoring function, which is responsible for these errors.
Note on Backwards Compatibility With DOCK 3.7
Previously it was said that SUBDOCK is compatible with DOCK 3.7- this is true, but with a caveat.
DB2 files generated for DOCK 3.8 will work in 3.7 via SUBDOCK, however they are known to produce spurious error messages. These can be ignored, for the most part, but may add some unwanted noise to your OUTDOCK file.
SUBDOCK help splash - all argument descriptions & defaults
[user@machine SUBDOCK]$ ./subdock.bash --help SUBDOCK! Run docking workloads via job controller of your choice =================required arguments================= expected env arg: EXPORT_DEST, --export-dest arg description: nfs output destination for OUTDOCK and test.mol2.gz files expected env arg: INPUT_SOURCE, --input-source arg description: nfs directory containing one or more .db2.tgz files OR a file containing a list of db2.tgz files expected env arg: DOCKFILES, --dockfiles arg description: nfs directory containing dock related files and INDOCK configuration for docking run expected env arg: DOCKEXEC, --dockexec arg description: nfs path to dock executable =================job controller settings================= optional env arg missing: USE_SLURM, --use-slurm arg description: use slurm defaulting to false optional env arg missing: USE_SLURM_ARGS, --use-slurm-args arg description: addtl arguments for SLURM sbatch command defaulting to optional env arg missing: USE_SGE, --use-sge arg description: use sge defaulting to false optional env arg missing: USE_SGE_ARGS, --use-sge-args arg description: addtl arguments for SGE qsub command defaulting to optional env arg missing: USE_PARALLEL, --use-parallel arg description: use GNU parallel defaulting to false optional env arg missing: USE_PARALLEL_ARGS, --use-parallel-args arg description: addtl arguments for GNU parallel command defaulting to =================input settings================= optional env arg missing: USE_DB2_TGZ, --use-db2-tgz arg description: dock db2.tgz tar files defaulting to true optional env arg missing: USE_DB2_TGZ_BATCH_SIZE, --use-db2-tgz-batch-size arg description: how many db2.tgz to evaluate per batch defaulting to 1 optional env arg missing: USE_DB2, --use-db2 arg description: dock db2.gz individual files defaulting to false optional env arg missing: USE_DB2_BATCH_SIZE, --use-db2-batch-size arg description: how many db2.gz to evaluate per batch defaulting to 100 =================addtl job configuration================= optional env arg missing: MAX_PARALLEL, --max-parallel arg description: max jobs allowed to run in parallel defaulting to -1 optional env arg missing: SHRTCACHE, --shrtcache arg description: temporary local storage for job files defaulting to /scratch optional env arg missing: LONGCACHE, --longcache arg description: longer term storage for files shared between jobs defaulting to /scratch =================miscellaneous================= optional env arg missing: SUBMIT_WAIT_TIME, --submit-wait-time arg description: how many seconds to wait before submitting defaulting to 5 optional env arg missing: USE_CACHED_SUBMIT_STATS, --use-cached-submit-stats arg description: only check completion for jobs submitted in the latest iteration. Faster re-submission, but will ignore jobs that have been manually reset defaulting to false