AFSample2

PaperCodeNeed >20GB

Modified AlphaFold for higher accuracy multimer structure prediction

Readiness

Readiness: Need >20GB
Smoke: passed
Artifacts: passed
Production: need_gt20gb_afsample2_alphafold_databases_or_precomputed_features
Storage: >20
Strategy: Stage kyogesh/afsample2 runtime as SIF plus full AlphaFold database tree, or explicitly constrain the tool to precomputed features/MSA inputs with validated examples.
Next: Install full AFsample2 SIF plus AlphaFold databases under shared storage, or redefine accepted inputs around precomputed features/MSA; then run a real Slurm benchmark and wire the adapter.

2026-07-17T13:13:16+08:00 SOP Need >5GB batch: reclassified to Need >20GB. Existing evidence shows AFsample2 source staged at /media/nik/seagate_nik/bio_server/tools/afsample2 with ~237MB code, but full production path requires AlphaFold-style databases (BFD, PDB70, PDB mmCIF, MGnify, UniRef90, UniRef30) unless the service is narrowed to precomputed features/MSA inputs. No completed SIF/database bundle or Slurm benchmark exists under current shared root.