DISCO
Co-design a protein sequence and structure.
Results
Loading starting example…
Method & evidence
Generate one paired 40-residue protein sequence and structure for early design exploration.
Scope & limitations
- Ligand-, DNA-, and RNA-conditioned modes remain unavailable until the exact expanded candidate passes final qualification.
Interpreting results
- Sequence and structure are generated together, so evaluate them as one design proposal.
- Inspect geometry and sequence liabilities before selecting a candidate for downstream modeling.
- Computational generation does not establish expression, stability, function, or safety.
References
Input requirements & outputs
Input checks
- Unconditional release: one 40-residue sequence–structure pair with seed 0
- Expanded release: 1–8 rows × 1–4 seeds with 30–500 designed residues per row
- Ligand rows require one V2000 SDF; DNA/RNA rows require explicit canonical sequences
- Use the result as a computational starting point, not experimental evidence
Expected results
- Designed sequence — Review the amino-acid sequence generated jointly with the structure.
- Designed structure — Open the paired PDB and inspect its fold and residue mapping.
- Conditioner and batch evidence — Review row order, seed identity, model-load count, and ligand or nucleic-acid presence evidence.
