Parameterization toolkits

Assigning atomic partial charges using some flavor of AM1-BCC with polymerist also requires installation of some supplementary toolkits.

One can mix-and-match installing any combination of the toolkits below to taste. If impatient or indifferent, one can opt for a “shotgun” approach and install all supported toolkits at once with the following commands:

mamba activate polymerist-env
mamba install -c conda-forge openff-nagl
mamba install -c openeye openeye-toolkits

OpenFF NAGL

This is an OpenFF-specific GNN based on similar architecture to Espaloma with a generally better validated partial charge model. To install individually, follow the install instructions provided by the OpenFF NAGL documentation.

OpenEye toolkits (closed-source)

These toolkits are required to perform explicit AM1-BCC charge assignment with conformer selection and enhance OpenFF’s conformer generation, but `polymerist` does not *require* OpenEye to work.

If you already have (or can obtain) an OpenEye license and wish to install the OpenEye toolkits individually, follow the install instructions provided by OpenEye.

Espaloma-charge (outdated, NOT recommended)

This is a standalone graph neural network (GNN) model [Wang et. al.] which can assign atomic partial charges trained on AM1-BCC data extremely rapidly. Due to lack of maintenance and dependency conflicts with Python 3.12, using espaloma-charge with polymerist is not longer encouraged, and support has been kept purely for backwards-compatiblity with legacy code. To install individually, follow the install instructions provided by Espaloma’s developers.