Nanobody Polyreactivity
Compare predicted nanobody polyreactivity.
Results
Loading starting example…
Method & evidence
Apply both published Harvey et al. CDR-only logistic models to heavy-chain nanobody variable domains and retain the exact model and numbering evidence.
Scope & limitations
- The first release scores sequences only and does not generate mutations or perform exhaustive rescue scans.
- It does not assess affinity, expression, aggregation, immunogenicity, clinical safety, or general antibody polyreactivity.
- Experimental polyspecificity testing remains required before candidate decisions.
Interpreting results
- Higher values indicate lower predicted polyreactivity; do not invert this direction.
- The values are relative logistic decision scores, not probabilities or calibrated pass/fail thresholds.
- Rankings outside synthetic camelid nanobody-like sequence space require especially cautious interpretation and experimental confirmation.
References
Input requirements & outputs
Input checks
- Canonical 20-amino-acid alphabet only
- 80–200 residues per sequence and at most 100 rows
- AbNumber must classify the variable domain as heavy chain under IMGT numbering
- CDR lengths must fit the fixed published one-hot feature profiles
Expected results
- Relative decision-score comparison — Review original-FACS and deep-FACS scores side by side; higher values indicate lower predicted polyreactivity.
- CDR and biochemical evidence — Inspect aligned CDRs, CDR lengths, isoelectric point, Kyte–Doolittle hydrophobicity, and N-glycosylation motif flags.
- Immutable execution evidence — Retain request, model-file, model-bundle, native-result, and content-hashed Artifact provenance.
