Home

Awesome

Please note: this repository is no longer being maintained.

License docs PyPI version Code style: black GitHub contributors Latest tag

Icolos: Workflow manager

The Icolos tool is a workflow manager for structure-based workflows in computational chemistry, that abstracts execution logic from implementation as much as possible. Icolos was designed to interface with REINVENT, and workflows can be called as a component of the scoring function, or to postprocess results with more expensive methods. Workflows are specified in JSON format (see folder examples). Currently wrapped are a diverse set of tools and internal steps, including docking, QM and MD capabilities. The pre-print is available here.

Introduction

Icolos provides a unified interface to a host of software for common computational chemistry calculations, with built in parallelization, and straight-forward extensibiltiy to add additional functionality. It was principally developed to handle structural calculations for REINVENT jobs, however, workflows can also be run independently.

Workflows are constructed from elementary 'steps', individual blocks which are combined to specify a sequential list of operations, with control of the command-line options provided through step settings, and options to control other aspects of the step's behaviour included in the additional block.

For many use cases, one of the template workflows might suit your needs, or need a few tweaks to do what you want. Demonstration notebooks for common workflows are available here.

Initial configuration

You are welcome to clone the repository and use a local version, and in particular if you would like to experiment with the code base and/or contribute features, please get in contact with us.

Installation

After cloning, first install and activate the icolosprod conda environment. To ensure the right installation directory is used, you can add the --prefix parameter to the create call, specifying the location of the conda environments.

conda env create -f environment_min.yml
conda activate icolosprod

Then install the package:

pip install -e .

This will give you access to the icolos entrypoint.

ESPsim installation

The following will install the ESPsim package into the environment - this is only required if ligand-based matching using this package is desired.

cd ..
git clone https://github.com/hesther/espsim.git
cd espsim
conda activate icolosprod
pip install -e .

Unit testing

Icolos is extensively unit tested, and relies on an external data repo located here. The full test suite takes ~60 mins on a workstation, therefore it is recommended that you execute a subset of unit tests relevant to the workflow you are running. To execute the full test suite, run something like:

pytest -n 8 tests/

Execution

Once a JSON is specified, the workflow can be executed like so:

conda activate icolosprod
icolos -conf workflow.json

We usually advise to check the validity of your configuration file before you try to execute it. There is a bespoke validator entry point to facilitate this:

validator -conf workflow.json

SLURM Execution

Once specified, a workflow can be called like this in a bash script:

#!/bin/bash -l
#SBATCH -N 1
#SBATCH -t 0-02:59:00
#SBATCH -p core
#SBATCH --ntasks-per-node=5
#SBATCH --mem-per-cpu=2G

source /<conda_path>/miniconda3/bin/activate /<conda_path>/minconda3/envs/icolosprod
icolos -conf workflow.json

For GROMACS workflows requiring the GPU partition, you will need to adapt the header accordingly, e.g. like so:

#!/bin/bash
#SBATCH -J gmx_cco1_fold_microsecond
#SBATCH -o MygpuJob_out_%j.txt
#SBATCH -e MygpuJob_err_%j.txt
#SBATCH -c 8
#SBATCH --gres=gpu:1
#SBATCH --mem-per-cpu=4g
#SBATCH -p gpu
#SBATCH --time=12:00:00

Documentation

The API reference is available here.

Developers