ESM-IF1
Inverse folding from backbone structures with ESM
ESM-IF1 is an inverse-folding model trained on 12 million CATH-derived and predicted structures to recover sequences for a given backbone. It captures sequence diversity by sampling from a learned distribution and scores candidate designs by likelihood. Useful for redesigning stable scaffolds and annotating designability of predicted structures.
ESM-IF1 is being onboarded — request access and be first in line.
At a glance
- Input
- Protein backbone (PDB)
- Output
- Designed sequences (FASTA)
- Developed by
- Meta AI
- Published
- Hsu et al., ICML 2022 · 2022
#inverse-folding#language-model#sequence-design
Related tools
More methods in Protein Design & Engineering
