Antibody Evolution

5-30 minutesPaperCodeNeed >5GB

Language models to recommend mutations for increased antibody binding affinity

Readiness

Readiness: Need >5GB
Smoke: passed
Artifacts: passed
Production: need_gt5gb_pytorch_esm_runtime_pending
Storage: 5-20GB
Strategy: large shared SIF/database/model bundle phase
Next: Create shared PyTorch/ESM runtime on symlink-capable storage or SIF/overlay, cache required ESM assets, then run minimal bin/recommend.py antibody benchmark.

No obvious large model/database dependency identified; validate upstream docs before install. Worker evidence 2026-07-01: source https://github.com/brianhie/efficient-evolution.git revision e9e8f610a5b4b3f48515741323209232faaf5039 staged at /media/nik/seagate_nik/bio_server/tools/antibody-evolution/efficient-evolution, clean checkout. Tool dir ~1.3GB mostly pip_cache. Slurm/GPU available but no job submitted. Code-level sanity: bin/recommend.py CLI takes wildtype sequence and ESM model names and emits mutation recommendations. Runtime benchmark not completed: venv on Seagate failed due lib64 symlink permission Operation not permitted; pip --target fallback began downloading large Python 3.12 CUDA/PyTorch wheels and was stopped. 2026-07-07 SOP reclassification: agent audit confirmed the blocker is the shared PyTorch/ESM runtime/model cache, not a small adapter-only issue. Moved to Need >5GB until the common ESM/PyTorch asset bundle is built and benchmarked. 2026-07-17T01:36:00+08:00 SOP Need >5GB next-10 audit: remains Need >5GB. Existing notes identify shared PyTorch/ESM runtime/model cache as blocker; current backend inventory did not find the prior antibody-evolution source path under /media/nik/seagate_nik. /home/nik has some ESM-capable environments, but no efficient-evolution repo/runtime benchmark and no adapter.