Effect-driven peptide design
Most platforms ask "which target?". We ask "what effect?". Describe the biology you want; the platform maps it to ranked targets, fires Phase 4 against each in parallel, and produces designed peptides with cryptographic provenance.
First-of-its-kind. Browse past runs →
How it works
- Your effect description is matched against the 42-target catalog by a semantic ranker.
- The ranker returns a JSON ranking with a one-sentence biological rationale per match.
- The top-N targets fire Phase 4 in parallel
(
RFdiffusion → ProteinMPNN → Boltz-2 → PoP-Shield → ClinicalGate → PDA). - Each run emits a cryptographic envelope hash binding the candidates to the pinned pipeline manifest.
- Candidates are persisted with sequences redacted under public scope; new code IDs (e.g.
PRX-MC4R-007) auto-increment. - A synthesis-agent finding is auto-posted to each affected fold linking back to this run.
- The run is browseable forever at
/by-effect/{id}with a Markdown export. - If the semantic ranker is unavailable, the system falls back to keyword overlap (still functional).