YAML Metadata Warning:empty or missing yaml metadata in repo card

Check out the documentation for more information.

hERG best model β€” DEPLOYMENT REFIT

DEPLOYMENT REFIT β€” trained on ALL labeled data including the benchmark sets (hERG-70, new-compounds, FDA panel, REGFAIL β€” absorbed 2026-11-14 by Hossain's decision). Benchmark metrics do NOT apply to this version; it has not been and must not be evaluated on the absorbed sets.

The honest, locked-pool benchmark of the winning recipes (models trained only on the decontaminated pools, external sets locked out, overlap re-verified = 0) is in benchmark_summary.json; the selection record is in selection.json.

Winners (by internal scaffold-split validation)

  • Classification: ensemble (val MCC 0.5729)
  • Regression: ensemble (val RMSE 0.6449)

Feature stack

Morgan ECFP4 bits + counts; RDKit 2D descriptors + QED + hERG-targeted physicochemical features; ChemBERTa-77M pretrained embeddings (mean-pooled, inference-only). CDK descriptors via PaDEL were attempted (time-boxed) and substituted by the RDKit 2D descriptor block after PaDEL stalled β€” see FEATURE_NOTES.md. No app/Space is provided.

Files

  • config.json β€” winner recipes, ensemble weights, training-set sizes.
  • tree_*.pkl / ltn_*.keras + ltn_*_preps.json β€” weights + fold preprocessing.
  • feature_extraction.py, predict_herg.py β€” end-to-end inference from SMILES.
  • benchmark_summary.json, selection.json, FEATURE_NOTES.md β€” provenance.
Downloads last month
66
Inference Providers NEW
This model isn't deployed by any Inference Provider. πŸ™‹ Ask for provider support