Summary
ML-based protein design methods outperform physics-based methods on identifying deleterious sequences but not top sequences (1). This compared ProteinMPNN, MIF-ST, ESM2, and Rosetta.
1.
Ertelt M, Moretti R, Meiler J, Schoeder CT. Self-supervised machine learning methods for protein design improve sampling but not the identification of high-fitness variants. Science Advances. 2025;11(7). Available from: https://doi.org/10.1126/sciadv.adr7338