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

Check out the documentation for more information.

hERG XGBoost β€” DEPLOYMENT REFITS

DEPLOYMENT REFIT β€” trained on ALL labeled data, including the regulatory/external sets (hERG-70, new-compounds 967, FDA panel 36, REGFAIL). These are the comprehensive-features XGBoost winners of the locked-pool Phase-5 benchmark (Approach A finals), refit on the complete labeled corpus for production screening use:

  • classification: 14,441 compounds (cls_alldata.csv)
  • regression: 10,370 compounds (reg_alldata.csv)

Hard gate: every external/regulatory compound was verified present in the actual training files by InChIKey connectivity before training (counts in validation_results.json).

Do not validate these models on the absorbed sets β€” scores there are in-sample by construction. Honest external benchmarks of the same architectures (trained with all external sets locked out) are in the benchmark repo Delower/herg-xgboost, which is untouched by this refit.

Contents

  • models/ β€” 4 bundles (cls/reg x scaffold/random; per task both protocols wrap the same fitted booster: identical combo/hyperparameters + all-data training make the protocol distinction vanish). Headline = scaffold.
  • predict_herg.py β€” standalone inference (SMILES -> blocker probability / pIC50).
  • feature_extraction.py β€” the exact Phase-5 feature blocks.
  • config.json, validation_results.json, regulatory_inference_sample.csv (in-sample demonstration predictions on FDA-panel compounds, labelled as such).

Hyperparameters (identical to the locked-pool finals): n_estimators 900, lr 0.03, max_depth 8, subsample 0.8, colsample_bytree 0.8, hist, seed 42. Classification cutoff: 10 uM (blocker = pIC50 >= 5.0; call = prob >= 0.5).

Downloads last month
37
Inference Providers NEW
This model isn't deployed by any Inference Provider. πŸ™‹ Ask for provider support