Inverse folding sequence perplexities correlate with Rosetta energies, forward folding TM-scores, and sequence recovery

Summary

Inverse folding perplexities/log-likelihood values correlate with energies calculated with Rosetta, TM-score of forward-folded decoys using OmegaFold (1), and sequence recovery (2).

Figures

Figure 4 from (1)

Figure 1 from (2)

See also

• Sequence recovery of inverse folding lower for smaller proteins

1.

Liu J, Guo Z, You H, Zhang C, Lai L. All‐Atom Protein Sequence Design Based on Geometric Deep Learning. Angewandte Chemie International Edition. 2024;63(50). Available from: https://doi.org/10.1002/anie.202411461

2.

Akpinaroglu D, Seki K, Guo A, Zhu E, Kelly MJS, Kortemme T. Structure-conditioned masked language models for protein sequence design generalize beyond the native sequence space. openRxiv; 2023. Available from: https://doi.org/10.1101/2023.12.15.571823