Inverse folding models trained on all proteins generate sequences that are equally thermostable as those generated by models trained only on thermophilic proteins

Summary

Inverse folding models trained on all proteins (including proteins from mesophiles) generate sequences that are as thermostable as those generated by models trained only on proteins from thermophiles ((1); $n=1$ warning). However, the sequence composition of the designs changes. This was observed using a custom-trained version of ProteinMPNN.

See also

• Thermophilic proteins have more aliphatic and charged residues on surface than mesophilic proteins

1.

Ertelt M, Schlegel P, Beining M, Kaysser L, Meiler J, Schoeder CT. HyperMPNN – A general strategy to design thermostable proteins learned from hyperthermophiles. openRxiv; 2024. Available from: https://doi.org/10.1101/2024.11.26.625397