QupKake

5-30 minutesPaperCodeProduction Ready

Predict pKa for organic molecules

Readiness

Readiness: Production Ready
Smoke: passed
Artifacts: passed
Production: ready
Storage: <5
Strategy: shared python312_slim SIF + QupKake Python site-packages + bundled xTB 6.4.1
Next: Production Ready for single-molecule SMILES pKa prediction. Optional hardening: add CSV/SDF batch upload controls and review whether to display acidic/basic pKa separately in the GUI.

SOP probe used upstream QupKake revision ceb4d99c2de3b9a314395c561db65a69a2b1392a at /media/nik/seagate_nik/bio_server/tools/qupkake. Env path /media/nik/seagate_nik/bio_server/envs/qupkake_sop_py312 is ~8.8 GiB; source ~699 MiB. Slurm job 360 completed acetic acid CC(=O)O benchmark, output SDF parsed with RDKit and predicted acidic pKa 4.940186, close to expected ~4.76; also emitted basic pKa 2.259321 requiring review. Upstream import needs conda/xtb workaround because no conda/mamba; XTBPATH points to bundled xtb. torch_geometric import cold-start ~179s. Real runner adapter/web artifact check not yet wired. 2026-07-07 18:10 +08:00 production validation: real QupKake runner adapter wired and deployed to Ubuntu backend. Runner API remote job fca33e30f7d2 / Slurm job 504 completed using shared python312_slim.sif, qupkake_sop_py312 site-packages, Palm RDKit fallback site, fake conda xtb-version shim, RDKit Draw stub for slim container, and bundled xTB at /media/nik/seagate_nik/bio_server/tools/qupkake/qupkake/xtb-641/bin/xtb. Acetic acid SMILES CC(=O)O produced qupkake_predictions.sdf, results.csv, qupkake_summary.csv, job.log, qupkake stdout/stderr, slurm.out, and slurm.err. Runtime 23.736s; predicted acidic pKa 4.867223 and basic pKa 2.246911.