Exploring atom mapping quality for predicting reaction barriers

This repo provides the necessary code and data to reproduce the results in the reply to comment on "Physics-based representations for machine learning properties of chemical reactions". This paper as well as the associated code explore the dependency of reaction barrier prediction to the atom-mapping quality.

Data

The datasets focussed on here are the Cyclo-23-TS, GDB7-22-TS and Proparg-21-TS. For each dataset, the following information is available

• The xyz structures of reactants and products, and computed barriers
• Atom maps for the original-mapped reactions (where available), RXNMapper reactions, and randomly mapped reactions.
• Corresponding submission csv files for chemprop
• The results of a 10-fold CV run of physics-based model SLATM+KRR

Codes

The codes in src/ provide:

• Means to atom map the reactions using RXNMapper, using the files mapper.py and process_maps.py
• Random mappings are generated using random_mapper.py
• The files learning.py and reaction_reps.py are support files for slatm_krr.py which performs the 10-fold CV prediction with 80/10/10 splits of the SLATM+KRR model

CGR results

Results of the CGR runs are saved to results/.

To run CGR models yourself, you will need to do so using chemprop, with the csv files in the data directory.