Protein Hunter

1-5 minutesPaperCodeNeed >5GB

De novo protein design

Readiness

Readiness: Need >5GB
Smoke: passed
Artifacts: passed
Production: deferred_full_model_runtime
Storage: 5-20
Strategy: shared PyTorch SIF + tool ext3 overlay; avoid per-tool torch duplication
Next: Build dedicated Protein Hunter SIF with Python 3.10, libgfortran5, PyTorch, Boltz/Chai dependencies, external model caches/weights, then run a real full-pipeline GPU smoke benchmark.

SOP probe used upstream Protein-Hunter revision a3e048e96c93953f022d15a4d35cd744b7e39e41 at /media/nik/seagate_nik/bio_server/tools/protein-hunter. Lightweight structural utility path installed under shared python312_slim SIF with target deps; Slurm job 353 completed and converted bundled CIF to PDB with sane two-chain metrics (A 89 CA, B 126 CA, Rg A 12.02 ?, Rg B 15.40 ?, centroid distance 21.41 ?). This is not full Protein Hunter production validation: PyTorch install timed out, Boltz/Chai weights were not downloaded, AF3 path not installed, and DAlphaBall.gcc lacks libgfortran.so.5 inside slim SIF.