Biohub Unveils World Model for Protein Biology

Biohub has unveiled the Protein Biology World Model, a scientific engine that can predict the structure of proteins and design new protein binders. This model is used to map proteins across the entire domain of life and predict their functions.

This release centers on three core technologies:

• ESMC: A state-of-the-art language model trained on approximately 2.8 billion sequences, internalizing the rules of protein folding, interaction, and function.
• ESMFold2: A design engine that converts ESMC's sequence representations into atomic-level 3D structures. It completes protein binder design for cancer and immunology-related targets—work that used to take months—in just days, while demonstrating high affinity and stability.
• ESM Atlas: The world's largest AI protein database, enabling exploration of 6.8 billion protein sequences and 1.1 billion predicted structures.

This model uncovers evolutionary connections that existing databases have failed to capture, helping researchers explore uncharted biological territory. All tools are provided free of charge to the global research community through the Biohub Platform.