BRER MD

Source: https://github.com/kassonlab/brer-md

Documentation: https://kassonlab.github.io/brer-md/

The brer-md Python package provides a set of scripts for running Bias-Resampling Ensemble Refinement (BRER) simulations using gmxapi. Details of the BRER method may be found in:

Hays, J. M., Cafiso, D. S., & Kasson, P. M. Hybrid Refinement of Heterogeneous Conformational Ensembles using Spectroscopic Data. The Journal of Physical Chemistry Letters. DOI: 10.1021/acs.jpclett.9b01407

Installation

Requirements

If you’re going to use a pip or a conda environment, you’ll need:

• Python 3.8 or newer.
• A GROMACS installation supporting client software builds.
• gmxapi for GROMACS.

brer-md includes a simple C++ extension module that can be attached to a GROMACS molecular dynamics (MD) simulator through the gmxapi Python interface. GROMACS installations (and GROMACS dependencies) can be built rather specifically for their computing environments. The brer package is distributed as source code that must be built for a specific GROMACS installation.

Note

For several recent versions of GROMACS, the “legacy API” needs to be enabled explicitly when GROMACS is configured. The GMX_INSTALL_LEGACY_API GROMACS CMake variable is not documented. Example:

cmake /path/to/gromacs/sources -DGMX_INSTALL_LEGACY_API=ON -DGMX_THREAD_MPI=ON

Python environment

We recommend using a separate Python virtual environment for each research project, tied to specific versions of the software tools you use. If your computing environment provides Python packages such as mpi4py that may be difficult to configure, you should try to use the provided packages in your virtual environment. Example:

python3 -m venv --system-site-packages myprojectvenv
. myprojectvenv/bin/activate
myprojectvenv/bin/python -m pip install --upgrade pip

Then, follow the installation instructions for GROMACS and gmxapi.

Build and Install

We recommend installing the package with pip.

Generally, pip will automatically install any package dependencies.

If a GROMACS installation is discoverable (you have "sourced" a GMXRC file or defined a GROMACS_DIR environment variable), then the gmxapi Python package will be installed automatically with the brer package. Simply:

pip install --pre brer-md
# or
pip install git+https://github.com/kassonlab/brer-md.git

If you prefer to install gmxapi separately (such as to specify an older package version), you can provide --no-deps and --no-build-isolation to pip install, and the existing gmxapi installation will be used.

You can pre-install (other) required packages using the requirements.txt file. The requirements.txt file does not include the gmxapi dependency.

Example:

pip show gmxapi |grep Version
# Version: 0.3.1
wget https://github.com/kassonlab/brer-md/blob/main/requirements.txt
pip install -r requirements.txt
pip install --no-deps --no-build-isolation brer

The Python package builder will manage compilation of the C++ GROMACS client using cmake documentation. If the GROMACS installation or C++ toolchain cannot be determined automatically, you may need to provide additional hints. See also GROMACS release notes.

Example:

gmx --version |grep prefix
# Data prefix:  /Users/eric/install/gromacs2022
CMAKE_ARGS="-C /Users/eric/install/gromacs2022/share/cmake/gromacs/gromacs-hints.cmake" \
pip install brer

Running BRER

Launching a single ensemble member.

An example script, run.py , is provided for ensemble simulations.

Let’s work through it piece by piece.

#!/usr/bin/env python

"""
Example run script
for BRER simulations
"""

import brer.run_config as rc
import sys

The import brer.run_config statement imports a RunConfig object, which handles the following things for a single ensemble member:

1. Initializing/setting up parameters for the BRER run.
2. Launching the run.

Then we provide some files and directory paths to the RunConfig object.

init = {
    'tpr': '/home/jennifer/Git/brer-md/tests/syx.tpr',
    'ensemble_dir': '/home/jennifer/test-brer',
    'ensemble_num': 5,
    'pairs_json': '/home/jennifer/Git/brer-md/tests/pair_data.json'
}

config = rc.RunConfig(**init)

In order to run a BRER simulation, we need to provide :

1. a tpr (compatible with GROMACS 2019).
2. The path to our ensemble. This directory should contain subdirectories of the form mem_<my ensemble number>
3. The ensemble number. This is an integer used to identify which ensemble member we are running and thus, the subdirectory in which we will be running our simulations.
4. The path to the DEER metadata. Please see the example json in this repository: src/brer/data/pair_data.json

Finally, we launch the run!

config.run()

You may change various parameters before launching the run using config.set(**kwargs) . For example:

config = rc.RunConfig(**init)
config.set(A=100)
config.run()

resets the energy constant A to 100 kcal/mol/nm^2 before launching a run.

Launching an ensemble

Right now, the way to launch an ensemble is to launch multiple jobs. We hope to soon use the gmxapi features that allow a user to launch many ensemble members in one job.

Troubleshooting

Mismatched compiler toolchain

One of the most common installation problems is related to incompatible compiler toolchains between GROMACS, gmxapi, and the plugin module. * CMake may warn “You are compiling with a different C++ compiler from the one that was used to compile GROMACS.” * When you import the brer module, you may get an error like the following. ImportError: dlopen(...): symbol not found in flat namespace '__ZN6gmxapi10MDWorkSpec9addModuleENSt3__110shared_ptrINS_8MDModuleEEE'

You can either set the CMAKE_CXX_COMPILER, explicitly, or you can use the GROMACS-installed CMake hints file.

You will have to rebuild and reinstall the brer module.

Remove any cached built packages:

pip cache remove brer

If you previously installed without build isolation you may have build or dist directories that should be removed, as well.

When attempting to build the package again, provide extra hints to CMake through the Python package builder by adding strings to the CMAKE_ARGS environment variable.

For GROMACS 2022 and newer, you would invoke cmake with something like the following. (The exact path will depend on your installation.)

CMAKE_ARGS="-C /path/to/gromacs_installation/share/cmake/gromacs/gromacs-hints.cmake" \
pip install brer

For GROMACS 2021 and older,

CMAKE_ARGS="-DCMAKE_TOOLCHAIN_FILE=/path/to/gromacs_installation/share/cmake/gromacs/gromacs-toolchain.cmake" \
pip install brer

See GROMACS release notes.

Problems building a GROMACS 2019 stack

For some C++ standard library installations, GROMACS 2019 exhibits compiler errors. The sources need to be patched. You can use the ci_scripts/limits.patch file in this repository as a guide to manually edit the source, or use the patch command line tool. Example:

cd /path/to/gromacs2019/sources
wget https://raw.githubusercontent.com/kassonlab/brer-md/main/ci_scripts/limits.patch
patch -p1 < limits.patch

For GROMACS 2019, you will need gmxapi 0.0.7. See https://gmxapi.readthedocs.io/en/release-0_0_7/.

You will have to prevent brer-md from trying to install a more recent version of gmxapi. Install the dependencies explicitly, then suppress automatic dependency resolution when installing brer-md. Exxample:

wget https://raw.githubusercontent.com/kassonlab/brer-md/main/requirements.txt
pip install -r requirements.txt
pip install --no-deps brer-md

References

Hays, J. M., Cafiso, D. S., & Kasson, P. M. Hybrid Refinement of Heterogeneous Conformational Ensembles using Spectroscopic Data. The Journal of Physical Chemistry Letters 2019. DOI: 10.1021/acs.jpclett.9b01407

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