Overview
Unlock the secrets of protein biology with ESMC, ESMFold2, and ESM Atlas—a revolutionary world model for prediction, design, and discovery. Built on cutting-edge Evolutionary Scale Modeling (ESM), this powerful system learns from billions of protein sequences, enabling unprecedented insights into protein structures, functions, and evolutionary relationships.
Key Features
- ESMC (Evolutionary Scale Model Chemistry): A state-of-the-art protein language model that captures the intricate rules of protein biology. ESMC excels in understanding long-range structural dependencies as model scale increases.
- ESMFold2: A leading protein structure prediction model that leverages ESMC embeddings and diffusion-based architecture. ESMFold2 accurately predicts high-resolution, all-atom 3D protein structures from sequences and has been validated for designing protein-protein interactions and novel protein binders.
- ESM Atlas: A comprehensive map of 6.8 billion proteins, featuring over one billion predicted structures. The Atlas is organized by ESMC's learned world model, made interpretable through sparse autoencoders that reveal ~16,000 functional features described in natural language.
Typical Use Cases
- Protein Design & Engineering: Generate novel protein sequences with desired properties, design specific protein-protein interactions, and engineer highly effective binders for therapeutic targets.
- Structure Prediction: Accurately predict the 3D structure of proteins, including complexes with DNA and ligands, from amino acid sequences.
- Bioinformatics Research: Explore evolutionary relationships, understand functional organization within proteins, and map the vast landscape of protein biodiversity.
- Drug Discovery: Accelerate the discovery of new therapeutic agents by designing and predicting the structure of novel protein binders.
