TEMPRO

secondsPaperCodeNeed >5GB

Nanobody melting temperature prediction

Readiness

Readiness: Need >5GB
Smoke: passed
Artifacts: passed
Production: deferred_esm_tensorflow_stack
Storage: 5-20
Strategy: shared python312_slim.sif + tool ext3 overlay
Next: Handle in 5-20GB round using shared ESM cache/container; convert TEMPRO.ipynb user workflow into CLI and validate sample.fasta or experimentals.fasta through Slurm with downloadable predictions.

SOP triage cloned upstream TEMPRO revision d2752834f16c30df1a684304a568ad1bac036583 at /media/nik/seagate_nik/bio_server/tools/tempro/source; checkout is 112M. user.zip includes small pretrained Keras models and TEMPRO.ipynb. Notebook workflow loads fair-esm ESM2-650M to generate sequence embeddings, then loads saved_ANNmodels_1500epoch/ESM_650M.keras. requirements pin Python 3.9.15, tensorflow 2.10.1, keras 2.10.0, torch 2.2.1, fair-esm. Production-valid run needs ESM model/cache plus old TensorFlow/PyTorch stack, expected to exceed <5GB unless a shared ESM/TensorFlow container is reused.