Minimization of Crystallography Open Database Crystal Structures with OpenFF

This repository contains code for downloading a set of crystal structures from the Crystallography Open Database (COD) (http://www.crystallography.net/cod/) and performing an energy minimization using the OpenMM (https://openmm.org/) molecular simulation tookit and OpenFF (https://openforcefield.org/).

Usage

Project was coded for Python 3.9. Refer to requirements.txt file for required packages.

Current Workflow

1. Download list of desired structures from COD using list of COD IDs in COD_ID_List.txt and COD_import.py script. This downloads CIF files from COD and converts to PDB format using pybel. These are output in the data/CIF and data/PDB directories.
2. Create supercell from PDB file using PyMol supercell function. This cell is created to be large enought to satisfy that periodic boundary conditions are greater than 1.8nm. This is done in the data/supercell_generation.py script and output in the data/PDB_supercell directory.
3. Paramaterize using openFF and build MD simulation in openMM. Perform energy minimization and record the initial and final energy values as well as the RMSD between initial and final state.
   1. main.py will perform an energy minimization which includes custom functions to minimize the unit cell as well as positions. This script will also output initial and final box vectors.
   2. main_no_box_minimization.py will perform only energy minimization with respect to position using the built-in OpenMM energy minimization.

Results that are output in the data directory:

1. minimzation_results.csv is a csv file with COD ID #, initial and final energies, and RMSD between those states. If box vector minimization is performed, the initial and final box vectors are also output.
2. minimization_results.pkl is a pickled file of the underlying Pandas DataFrame for the above data.
3. rmsd_values.txt is a tab separated text data file that only reports COD ID and RMSD values.
4. initial_states and final_states directories contain the .xml state files from OpenMM saved before and after energy minimzation.
5. dminimized_PDB_supercell directory contains the PDB files of the minimized supercell system.

Examples

The examples folder contains scripts for a reduced subset of 10 COD ID values with the expected outputs stored in the expected_data directory. For all, ensure the working directory is set to /examples.

1. Run example_COD_import.py.
2. Run example_supercell_generation.py.
3. Run example_main.py.
4. Compare the generated data in example_data to expected_data.

Current Issues

COD_ID_List.txt is not curated of entries that have data issues (non-matching coordinates between CIF and SMILES and entries without SMILES strings). These entries currently produce errors that are logged in errors.log during execution.

The current energy minimization algorithm fails for many systems due to periodic cell conditions not being in a reduced format (http://docs.openmm.org/7.0.0/userguide/theory.html#periodic-boundary-conditions).

The energy minimization algorithm also rarely tries to reduce the periodic boundary conditions to less than 0.9 nm which is half the non-bonded cutoff distance which causes an error.