BindCraft Pipeline
Computational protein binder design pipeline — from target selection through structure prediction to binding affinity scoring.
BindCraft automates the pipeline from target protein identification through binder candidate generation and scoring. Built on ESM-2 embeddings and Rosetta energy functions, deployed on federated GPU infrastructure.
The immediate target: BAFF (B-cell activating factor) miniprotein binders for Sjögren’s syndrome. The broader ambition: a disease-graph approach connecting proteins, pathways, and clinical endpoints across the rare disease landscape — eighteen targets cataloged so far, twenty-five dataset structures mapped.
The pipeline uses RFDiffusion for structure generation and ProteinMPNN for sequence design. Think of it as a conversation between two AI systems: one imagines what a protein binder might look like, the other figures out the amino acid sequence that could fold into that shape. The candidates get scored by Rosetta’s energy functions, and the best ones move forward for experimental validation.
None of this works without compute. BindCraft runs on Lattice Protocol’s federated GPU infrastructure — the same commitment that compute should never be rare for rare disease researchers, applied one binder at a time.