Nanobody Polyreactivity

New5-30 minutesPaperCodeProduction Ready

Nanobody (VHH) polyreactivity prediction

Readiness

Readiness: Production Ready
Smoke: passed
Artifacts: verified
Production: production_validated_runner_api_slurm_nanobody_polyreactivity
Storage: <5GB
Strategy: shared python312_slim.sif plus existing shared Torch/Pandas/Scikit-learn targets; optional future ANARCI wrapper for raw FASTA preprocessing
Next: Production ready for ANARCI/IMGT heavy-chain CSV scoring; optional future work is packaging ANARCI for raw FASTA preprocessing.

SOP agent cloned upstream nanobody-polyreactivity revision 1e8c720cff7c393b5976bbd6a678a691e8b04c16 to /media/nik/seagate_nik/bio_server/repos/nanobody-polyreactivity. App is Docker Compose/FastAPI with bundled model weights and examples. Minimal env install did not complete because Seagate rejects venv symlinks and pip downloads broke. Upstream hard-codes /nanobody-polyreactivity and /opt/conda/bin/ANARCI, so it needs a container/bind wrapper or proper image/env layout. No Slurm benchmark submitted. 2026-07-09T17:32:50+08:00 SOP production validation: promoted from Production Blocked to Production Ready. Validated upstream nanobody-polyreactivity logistic polyreactivity scoring using the bundled ANARCI/IMGT heavy-chain CSV and bundled model artifacts under /media/nik/seagate_nik/bio_server/repos/nanobody-polyreactivity/backend. Direct Slurm benchmark job 629 completed exit 0 inside shared python312_slim.sif with shared Torch/Pandas/Scikit-learn targets and scored 2 sequences with four models: origFACS_lr_onehot, origFACS_lr_3mer, deepFACS_lr_onehot, and deepFACS_lr_3mer. Implemented production runner adapter for ANARCI/IMGT heavy-chain CSV scoring and validated through Ubuntu runner API remote job b355d6cd6fd2 / Slurm job 633. The API run completed in 19.172 s, scored 2 sequences, produced origFACS one-hot score range -2.5742294725218113 to 0.3392225177775981, deepFACS one-hot score range -2.522768923169517 to 1.148063013497073, and predicted 1 polyreactive sequence by both origFACS and deepFACS one-hot models. Artifacts verified: summary.json, results.csv, sample_sequences_H.csv, nanobody_polyreactivity_scores.csv, nanobody_polyreactivity_summary.json, nanobody_polyreactivity_result.json, nanobody_polyreactivity_stdout.txt, nanobody_polyreactivity_stderr.txt, run_nanobody_polyreactivity.py, slurm.out, and slurm.err. Raw FASTA-to-ANARCI preprocessing remains a future optional packaging enhancement; production runner accepts validated ANARCI/IMGT heavy-chain CSV input.