ESM-2
Protein language model embeddings and scoring
ESM-2 is a large protein language model trained on evolutionary sequences that produces per-residue and sequence-level embeddings for downstream prediction tasks. It scores mutations and estimates secondary structure and contact maps directly from unsupervised attention maps. A foundational model for protein ML pipelines.
ESM-2 is live on MolPilot — paste a sequence and press Fold.
At a glance
- Input
- Protein sequence (FASTA)
- Output
- Embeddings and scores (NPZ)
- Developed by
- Meta AI
- Published
- Lin et al., Science 2023 · 2022
#language-model#embeddings#mutation-scoring
Related tools
More methods in Sequence Analysis & Alignment
