DeepSP

secondsPaperCodeProduction Ready

Viscosity, spatial charge map (SCM) and spatial aggregation propensity (SAP) predictions for antibodies

Readiness

Readiness: Production Ready
Smoke: passed
Artifacts: passed
Production: ready
Storage: 5-20GB
Strategy: shared python312_slim.sif + deepsp_py312_site TensorFlow/Keras CPU layer + palm_py312_site scientific dependencies
Next: Production Ready for pre-aligned antibody DeepSP descriptor inference. Optional hardening: add ANARCII raw-sequence numbering/alignment pipeline before enabling arbitrary unaligned antibody inputs.

SOP agent validated upstream DeepSP revision 86cd81dc22f003ff6dc0b0739f9a8c1a9727c953 at /media/nik/seagate_nik/bio_server/repos/deepsp. Repo is ~79M but dedicated TensorFlow target site is ~6.3G at /media/nik/seagate_nik/bio_server/envs/deepsp_py312_site. Used shared python312_slim.sif plus deepsp_py312_site TensorFlow CPU 2.18.1/Keras 3.8/h5py/NumPy and palm_py312_site for pandas/Biopython/sklearn. Slurm job 381 completed in 6.89s on bundled aligned input: 16 antibody sequences length 272, descriptor table 16x31, all numeric descriptors finite. Artifacts: /media/nik/seagate_nik/bio_server/tools/deepsp/validation_20260701_225910/. Prior job 380 captured Keras legacy deserialization issue. Upstream CLI is stale and uses legacy ANARCI; successful probe bypassed numbering with bundled pre-aligned input. 2026-07-07T21:09:06+08:00 SOP production validation: added real runner adapter for deepsp scoped pre-aligned antibody descriptor mode. Deployed via shared python312_slim.sif with deepsp_py312_site TensorFlow/Keras and palm_py312_site dependencies. Runner API job d49fbef525be / Slurm job 513 completed in 6.567 s using bundled seq_aligned_HL2 fixture, scoring 16 aligned antibody sequences length 272. Descriptor table shape was 16x31 with all finite numeric values and first descriptor sum 8218.048828125. Artifacts verified: summary.json, results.csv, deepsp_descriptors.csv, deepsp_runtime.json, input_aligned.txt, job.log, slurm.out, slurm.err. Production scope is pre-aligned DeepSP descriptor inference; raw sequence ANARCII numbering/alignment remains optional future hardening.