Kidney toxicity β€” Arch3 LazyPredict-best deployment model (BaggingClassifier)

Binary classifier for kidney toxicity (nephrotoxicity) concern: the probability that a compound belongs to the toxic class. This is the deployment refit of the kidney programme's Arch3 arm (the LazyPredict sweep), one of the two winners selected on the locked regulatory benchmark. It is deployed as a pair with the Arch5 deployment model (Delower/kidney-arch5-ensemble-deployment); a compound flagged by both models is the highest-concern readout.

Recipe

The arm's own procedure, unchanged, with no re-tuning beyond the sweep itself: a LazyPredict LazyClassifier sweep over the fp_desc feature block, ranked by validation MCC on an internal seed-42 scaffold split of the deployment pool, the probability-support guard applied, and the winning class refit on the full pool with DescPrep refit on the full pool.

  • The deployment sweep's winning class is BaggingClassifier (scikit-learn defaults as instantiated by the arm: 10 bootstrap ensembles of the default decision-tree base estimator, random_state=42). On the internal stage split it scored val MCC 0.282 / AUROC 0.695 (train-fold instance) β€” an internal selection signal on absorbed data, not a benchmark score.
  • Features (fp_desc, 1,402 columns): Morgan fingerprint bits (1024) + MACCS keys (167) + RDKit descriptors prepared by the arm's DescPrep (211 of 217 kept; preprocessing shipped inside the weights bundle).
  • Flag threshold: probability β‰₯ 0.5 (the programme convention). Bagging probabilities are ensemble vote fractions, hence the 0.1-step values seen in the screen output.

One naming point, stated exactly: in the locked wave, this arm's sweep winner was NearestCentroid β€” that is the model behind the benchmark scores below. The deployment refit re-ran the same sweep procedure on the deployment pool (which includes the absorbed benchmark), and there the winner is BaggingClassifier. Same arm, same procedure, different pool, different winning class.

Deployment training set

All labeled kidney data the programme holds, assembled by InChIKey connectivity dedup of the standardized parent with benchmark labels governing conflicts: 1,805 unique compounds (915 positive / 890 negative) = corrected regulatory benchmarks (169 kept) β†’ v1 benchmarks (0 new β€” the corrected set is a superset) β†’ internal test (327) β†’ enriched training pool (1,309) β†’ locked training set (0 new β€” fully covered). 1,307 duplicate rows dropped; 0 label conflicts. In short: locked training + internal test + enriched additions + every regulatory benchmark compound absorbed.

Hard gate (passed inside the training scripts before any training): every regulatory benchmark compound verified compound by compound inside the actual training matrix β€” corrected benchmark: withdrawn 16/16, clinical failures 7/7, PASS 146/146; original v1 benchmark: withdrawn 12/12, clinical failures 5/5, PASS 146/146; missing = 0 in every set. One exception, stated explicitly: each withdrawn file contains one further substance (17th corrected / 13th v1) with no resolvable structure β€” it cannot enter any feature matrix, was left unscored by every arm in every wave, and is excluded from the gate denominators for that reason. See gate_verification.json.

Honest benchmark scores β€” benchmark version only

The deployment refit absorbed the regulatory benchmark and is never scored on it. The frozen claim for this arm comes from the locked-wave benchmark model (NearestCentroid; benchmark locked out of training, lockout verified 0; v1 pooled basis, 163 scored compounds: 12 withdrawn + 5 clinical failures + 146 PASS):

Metric Value
Pooled regulatory AUROC 0.568
Pooled balanced accuracy 0.620
Withdrawn drugs caught 12/12
Clinical failures caught 5/5
PASS compounds correct 35/146 (0.240)

This arm's value is its balance: perfect withdrawn/failure recall with the best pooled balanced accuracy of any single kidney arm β€” the tempering half of the deployment pair.

Deployment screen results

Applied to the DPP-4 programme's potent leads (flag = prob β‰₯ 0.5):

  • 1,113 workbook leads β€” scored by both models 1,113/1,113: this model flagged 626 (56.2%), Arch5 flagged 668 (60.0%), flagged by both 419, by either 875, pass both 238.
  • 4,898 full potent union β€” scored by both models 4,898/4,898: this model flagged 3,041 (62.1%), Arch5 flagged 3,219 (65.7%), flagged by both 2,195, by either 4,065, pass both 833.

Flag rates of 56–66% are this programme's known high-sensitivity / low-specificity profile, stated plainly: every kidney arm trades specificity for withdrawn-drug recall, so a single-model flag is low-information on its own; the dual-flag column is the readout.

Files

  • model.joblib β€” joblib bundle: the fitted BaggingClassifier, the fitted DescPrep preprocessing object, and the winning class name. Unpickling requires deployment_common.py (shipped here) to be importable.
  • config.json β€” recipe, feature block, training composition, gate counts, benchmark scores.
  • deployment_common.py β€” the programme's shared featurization module (fingerprints, descriptors, DescPrep).
  • predict.py β€” prediction entry point (predict(smiles) / predict_many([...])); featurizes exactly as the deployment screen did.
  • validate.py + examples.json β€” runnable validation: reproduces this model's probabilities on 8 screened DPP-4 leads (tolerance 1e-6; the model is deterministic). Verified locally against this exact layout before upload: max deviation 0.
  • gate_verification.json, training_manifest.json β€” hard-gate result and full training manifest.
  • absorbed_compounds.csv β€” the 1,805-compound deployment training set (the absorbed data, for audit).

Limitations

Requires scikit-learn, rdkit, joblib, numpy. Probabilities are screening scores (ensemble vote fractions), not calibrated clinical risks.

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