Documentation | PyPI | GitHub

PyMPFIT is a free, open-source software for performing partial atomic charge fitting using the Gaussian distributed multipole analysis (GDMA). Features include:

• Built on open-source libraries: OpenFF Recharge, OpenFF Toolkit, and NumPy
• Direct interface to Psi4 / GDMA and from wavefunctions stored within QCFractal (i.e., QCArchive)
• Generating multipole moments for multi-conformer molecules
• An SQLite database backend for efficient high-throughput scaling
• Bayesian methods (Pyro) for flexible virtual site fitting

Installation

Dependencies

PyMPFIT depends on OpenFF packages and Psi4/GDMA, which are distributed via conda-forge. Works with conda, mamba, or micromamba.

conda create -n pympfit python=3.12 openff-recharge openff-interchange openff-utilities psi4 libint=2.9 pygdma -c conda-forge -y
conda activate pympfit

Note: The libint=2.9 pin is a temporary workaround for a psi4/libint2 compatibility issue. This will be resolved in psi4 v1.10 build 3.

Install

pip install pympfit

Optional: Bayesian Virtual Site Fitting

pip install pyro-ppl arviz matplotlib sphericart-torch

Development Installation

git clone https://github.com/shehan807/pympfit.git
cd pympfit
pip install -e ".[test]"

Quick Example

from openff.toolkit import Molecule
from openff.recharge.utilities.molecule import extract_conformers
from pympfit import (
    GDMASettings, Psi4GDMAGenerator, MoleculeGDMARecord,
    MPFITSVDSolver, generate_mpfit_charge_parameter,
)

# Create molecule
molecule = Molecule.from_smiles("CCO")
molecule.generate_conformers(n_conformers=1)
[conformer] = extract_conformers(molecule)

# Compute multipoles via Psi4/GDMA
settings = GDMASettings(method="pbe0", basis="def2-SVP", limit=4)
coords, multipoles = Psi4GDMAGenerator.generate(
    molecule, conformer, settings, minimize=True
)

# Fit partial charges to multipoles
record = MoleculeGDMARecord.from_molecule(molecule, coords, multipoles, settings)
charges = generate_mpfit_charge_parameter([record], MPFITSVDSolver())

import numpy as np

print(f"MPFIT SMILES         : {charges.smiles}")
print(f"MPFIT CHARGES        : {np.round(charges.value, 4)}")

Output:

MPFIT SMILES         : [H:1][O:2][C:3]([H:4])([H:5])[C:6]([H:7])([H:8])[H:9]
MPFIT CHARGES        : [ 0.34    0.6064 -0.6405 -0.1101 -0.1218 -0.113  -0.1281 -0.1534  0.3205]

See examples/tutorials/quickstart.py for more.

License

The main package is released under the MIT license.

Copyright

Copyright (c) 2026, Shehan M. Parmar

Acknowledgements

Project based on the Computational Molecular Science Python Cookiecutter version 1.11.