Zero-shot protein stability prediction using inverse folding models can be improved by subtracting predictions from residue in isolation

Summary

Zero-shot protein stability prediction using inverse folding models can be improved by subtracting predictions from residue in isolation (1). This corrects for the contribution of unique backbone geometries (particularly for glycine, proline, valine, and isoleucine) on predictions. It is equivalent to corrections made by FEP.

See also

• PLMs downweigh probability of sequences with multiple mutations
• Inverse folding models can predict antibody-antigen binding affinities with higher accuracy by correcting with predictions from unbound state

1.

Dutton O, Bottaro S, Invernizzi M, Redl I, Chung A, Hoffmann F, et al. Improving Inverse Folding models at Protein Stability Prediction without additional Training or Data. openRxiv; 2024. Available from: https://doi.org/10.1101/2024.06.15.599145