AF-Traj

30-60 minutesPaperCodeNeed >20GB

Predict protein conformations with subsampled AlphaFold2

Readiness

Readiness: Need >20GB
Smoke: passed
Artifacts: passed
Production: deferred_af2_colabfold_msa_database_runtime
Storage: >20GB
Strategy: shared python312_slim.sif + tool ext3 overlay
Next: Decide AF-Traj storage/runtime strategy: shared ColabFold/AF2 service with remote/public MSA for lighter mode, or full local databases in large-storage phase. Then stage notebooks/scripts as a thin adapter and run tiny monomer ensemble smoke test through Slurm.

SOP recovery agent triaged AF-Traj. Catalog source is GMdSilva/gms_natcomms_1705932980_data for Nature Communications 2024; upstream is notebooks/data for subsampled AlphaFold2 conformational ensemble prediction, not a packaged CLI. Staged path /media/nik/seagate_nik/bio_server/tools/af-traj exists but is effectively empty. Requirements depend on AF2/ColabFold-style prediction, MSA generation, JAX/AlphaFold/ColabFold, trajectory analysis packages such as MDTraj/MDAnalysis, and potentially local search databases up to ~3TB. RTX 5050 Laptop has about 8GB VRAM, which is risky for production AF2 ensemble runs. No Slurm benchmark run because source/runtime are absent.