FoldCraft
Redesign a binder against a fixed protein target.
Results
Loading starting example…
Method & evidence
Run FoldCraft against checksum-bound binder and target templates, then inspect every designed sequence, predicted complex, and stage-level score.
Scope & limitations
- The released batch reuses one ProteinMPNN load; upstream AlphaFold models reinitialize by stage and trajectory.
- Every output remains linked to both exact input PDBs.
Interpreting results
- FoldCraft scores are model-derived design evidence, not measured affinity, stability, expression, or function.
- Compare candidates only within runs using identical templates, chains, hotspots, models, and stage settings.
- Inspect interface geometry and sequence liabilities before advancing a design.
References
Input requirements & outputs
Input checks
- Exactly one checksum-verified binder PDB and one target PDB
- Explicit binder and target chains with valid target hotspot residue ranges
- One to four trajectories in the released batch, with up to four ProteinMPNN samples each
Expected results
- Designed sequences and ranked metrics — Compare trajectory, sequence, confidence, loss, and interface evidence in one table.
- Predicted binder–target structures — Open each checksum-bound PDB in the shared 3D structure viewer.
- Execution provenance — Retain templates, hotspots, models, parameters, image, request, result, and artifact digests.
