TAFFI-gen

Topology Automated Force-Field Interactions (TAFFI) is a framework for parameterizing transferable force-fields to quantum chemistry data. Here we are distributing force-field parameters for TAFFI-gen, a fixed charge, harmonic force field fit to wB97X-D3/def2-TZVP data with coverage for small organic molecules. The current distribution provides the TAFFI-gen force-field parameters used for the simulations reported here.

What's Included

The included scripts are provided as stand-alone python programs that only provide atomtyping and force-field assignment capability, not the full TAFFI functionality required for extending the force-field or parameterizing alternative functional forms.

• taffi_lite_tester.py: this is a driver script that will generate lammps input files for the benchmarked taffi-gen structures. This script calls the others and should be the first thing that the beginning user consults and tries to run.

• gen_md_for_sampling.py: this python script performs taffi-gen atom typing, force-field parameter assignment, and lammps input file generation.

• Lib: this folder contains some TAFFI library functions used by gen_md_for_sampling.py

• taffi_gen.db: this is a text file that contains the taffi-gen force-field parameters required to simulate the benchmarked structures.

• xyzs: this folder contains the *.xyz files for each benchmark molecule.

• OKKJLVBELUTLKV-UHFFFAOYSA-N: this folder contains quantum chemistry training data for the methanol model compound. This is included here as an example of the training data for interested readers of the TAFFI-gen reference. A full description of model compound and training data generation are supplied there.

Usage

Dependencies

• Python 3
• numpy
• openbabel 2.3.2 and above

Commands

• To create lammps input files using your xyz, run

python gen_md_for_sampling.py example.xyz

where example.xyz is a user-supplied structure, or one of the structures included in the "xyzs" folder. The output of this script will be the inputes for a lammps job. Optional arguments to the gen_md_for_sampling.py script can be reviewed by running it with the -h option. Note that the supplied taffi_gen.db file only has guarranteed coverage for the benchmarked model compounds. If the user supplies a compound with missing parameters, the script will print out the missing paramters for the user.

• To create input files based on the included benchmark .xyz files inside of the xyz folder run:

python taffi_lite_tester.py

Running this script will react a folder called test_runs that contains the input files for each job. This folder already exists in the repository for the user to check the expected behavior. If running from scratch, the existing test_runs folder should be renamed before running the script.

Citation

Please cite the following if you use TAFFI-gen

Bumjoon Seo, Zih-Yu Lin, Qiyuan Zhao, Michael A. Webb, Brett M. Savoie. Topology Automated Force-Field Interactions (TAFFI): A Framework for Developing Transferable Force-Fields. ChemRxiv. https://doi.org/10.26434/chemrxiv.14527299.v1