Biohub launches ESM protein world model
Biohub released ESM, an open protein-biology platform built from three parts: ESMC, ESMFold2, and ESM Atlas. The system learns from billions of protein sequences to model sequence, structure, and function, with ESMFold2 for structure prediction and binder design, ESMC as the foundation model, and ESM Atlas as a navigable map of billions of sequences and predicted structures. The release is positioned as a major step toward a world model of protein biology for research and therapeutic discovery.
Big scientific launch, not a typical startup product. The important shift is from single-task protein prediction to a shared model ecosystem that can support structure, function, and design workflows.
- –ESMC is the foundation model trained on billions of sequences, so the release is really about representation learning at biology scale.
- –ESMFold2 appears to be the most immediately practical piece for protein structure prediction and binder design.
- –ESM Atlas adds a searchable map of protein space, which is useful for discovery and hypothesis generation, not just prediction.
- –This is highly relevant to biotech and drug discovery teams, but less relevant to general AI audiences.
DISCOVERED
45d ago
2026-05-28
PUBLISHED
46d ago
2026-05-27
RELEVANCE
AUTHOR
SandboxAQ