BoltzGen
Generate target-conditioned binder candidates.
Results
Loading starting example…
Method & evidence
Generate a bounded protein or peptide binder cohort against a checksum-bound target structure and explicit epitope residues, retaining every native metric and filter decision.
Scope & limitations
- Fixed-seed exact replay is qualified on the released H100 image.
- Generated candidates require independent structural, developability, and experimental validation.
Interpreting results
- A native filter failure is a scientific result for that candidate, not an infrastructure failure.
- Use metrics to prioritize within a consistently configured target campaign; they are not experimental affinities.
- Campaign-scale success should be judged across candidates and followed by independent scoring and experiments.
References
Input requirements & outputs
Input checks
- One checksum-verified PDB or mmCIF target structure
- Explicit target chain and 1–32 epitope residue numbers
- Protein or peptide protocol with 1–4 designs per governed Run
Expected results
- Candidate cohort — Compare designed sequences, final ranks, native metrics, and filter outcomes.
- Designed complexes — Inspect each checksum-bound target–binder mmCIF structure.
- Native campaign evidence — Retain the exact image, source, models, seed, request, result, and structure digests.
