Skip to content

Gradient based analysis of force field optimizations.

License

Notifications You must be signed in to change notification settings

SimonBoothroyd/graffan

Repository files navigation

Gradient Based Force Field Analysis (GraFFan)

tests Language grade: Python codecov License: MIT

This framework aims to offer a set of common utilities for analyzing the outputs of force field optimization (performed by ForceBalance), with a focus on providing tools for inspecting the many contributions to the gradient of the objective function.

Installation

The package and its dependencies can be installed using the conda package manager:

conda env create --name graffan --file devtools/conda-envs/test_env.yaml
python setup.py develop

If any targets require MOL2 or GZipped SDF (*.sdf.gz) molecule files then you will also need to install the OpenEye toolkit as the openff-toolkit package does not currently support loading these formats through RDKit. The OpenEye toolkits can be installed by first activating the graffan conda environment created above, and then running:

conda install -c openeye openeye-toolkits

Getting Started

The framework offers two CLI utilities which encapsulate most of this frameworks features: graffan analyse and graffan visualise FILENAME.

The graffan analyse command should be run in the root directory of a force balance optimization and will analyze the outputs of each fitting target used in the optimization. Namely, it will create a new iteration_0000.json file (or similar depending on whether the --iteration X flag was used) which contains the contributions of each target to the total gradient of the objective function with respect to the force field parameters which were refit.

The graffan visualise iteration_0000.json command will then open up of GUI in a webbrowser allowing the extracted gradients to be viewed in higher detail.

Copyright

Copyright (c) 2020, Simon Boothroyd

About

Gradient based analysis of force field optimizations.

Resources

License

Code of conduct

Stars

Watchers

Forks

Packages

No packages published

Languages