alterlab-esm

What problem does it solve?

Researchers need a unified toolkit to design, predict, and analyze proteins using cutting-edge multimodal models. This Skill provides an integrated approach to generate sequences, predict structures, generate embeddings, and condition outputs on functional attributes, enabling rapid exploration of protein designs.

Core Features & Use Cases

• Multimodal design across sequence, structure, and function using ESM3 and ESM C models.
• Structure prediction and inverse folding to design sequences that fold into target structures.
• Protein embeddings for downstream ML tasks, clustering, and similarity analyses.
• Function conditioning and chain-of-thought generation for iterative protein design workflows.
• Local model usage and Forge API support for scalable inference and batch processing.

Quick Start

Provide a partial protein sequence (with masked regions) and run a basic generation to obtain a completed design.