DSMBind

5-30 minutesPaperCodeNeed >5GB

Predict antibody-antigen binding affinity

Readiness

Readiness: Need >5GB
Smoke: not_tested
Artifacts: not_applicable
Production: deferred_torch_esm_chemprop_rdkit_data_runtime
Storage: 5-20GB
Strategy: compiled/binary SIF or shared python312_slim + ext3 overlay
Next: Build shared PyTorch/ESM/Chemprop/RDKit/sidechainnet/SRU runtime, create CLI wrapper from tutorial notebook with a small bundled example, run Slurm benchmark, then add API artifacts.

SOP main-thread triage cloned upstream wengong-jin/DSMBind revision 66a3877b758b7ede9eb60110ee2be530345563d2 at /media/nik/seagate_nik/bio_server/tools/dsmbind/source; source about 197M with bundled ckpts including cross.ckpt, model.antibody.allatom, model.drug.allatom, model.skempi.allatom. README requires PyTorch 1.13, biotite, SRU++ 3.0.0-dev, ESM-2, Chemprop, tqdm, RDKit, sidechainnet <=0.7.6, and optional Zenodo training/test data. Upstream inference is notebook-based. No Slurm benchmark run because compatible runtime/data/notebook wrapper not staged.