The gmxapi project provides interfaces for managing and extending molecular
dynamics simulation workflows. In this repository, a Python package provides the
gmx
module for high-level interaction with GROMACS. gmx.core
provides
Python bindings to the gmxapi
C++ GROMACS external API.
The project is hosted on GitHub and
includes the gmxapi
repository along with supporting respositories. The
gromacs-gmxapi
repository
includes a modified version of GROMACS that supports the latest gmxapi
features not yet available through an official GROMACS distribution. A
sample repository illustrates
how to implement a GROMACS plugin that applies restrained-ensemble forces
to a loosely-coupled ensemble of simulation trajectories.
The whole thing is driven at the highest level by a simple Python interface.
See the latest documentation at http://gmxapi.readthedocs.io/
Irrgang, M. E., Hays, J. M., & Kasson, P. M. gmxapi: a high-level interface for advanced control and extension of molecular dynamics simulations. Bioinformatics 2018. DOI: 10.1093/bioinformatics/bty484
Note: automated documentation generation may not be up to date for development branch.
To run a simulation as you would with the gmx
command-line tool, a gmxapi
Python script would look simply like the following.
import gmx
md = gmx.workflow.from_tpr('myTPRfile.tpr')
gmx.run(md)
With a plugin, such as the sample ensemble-restraint, you can
run a coupled ensemble of simulations that collectively refine a bias potential.
Assuming tpr_list
is a Python list of input files defining the ensemble, and
params
is a Python dictionary of keyword arguments,
import gmx
import myplugin
md = gmx.workflow.from_tpr(tpr_list)
potential = gmx.workflow.WorkElement(
namespace="myplugin",
operation="ensemble_restraint",
depends=[],
params=params)
potential.name = "ensemble_restraint_1"
md.add_dependency(potential)
gmx.run(md)
Currently, given an MPI environment, gmxapi would run the above workflow on a
number of nodes equal to the length of the tpr_list
array, parallelizing each
simulation across a single node, periodically sharing data via MPI calls across
the ensemble.
Many more features and more flexibility are still to come. Feel free to make suggestions or describe your own priorities and research needs in the issue tracking system.
A compatible version of GROMACS must be installed before building the Python
package. See gromacs-gmxapi
or GROMACS 2019.
We recommend installing gmxapi in a Python virtual environment. See docs/install.rst for details or refer to the online user documentation.