Genotoxicity Predictor
Predicts Ames mutagenicity from molecular structure (SMILES) using
ECFP4 fingerprints and a random-forest classifier — a computational companion to the
RAD52‑GFP DNA‑damage biosensor.
Model loaded · 108 compounds
Screen a compound
Type a chemical name (e.g. aspirin, nicotine) or paste a SMILES — or click an example above.
Prediction
Confidence: · canonical
Non-mutagenicP(mutagenic)Mutagenic
0% probability mutagenic
Most similar training compounds
Model performance
0.998
ROC‑AUC
96.3%
Accuracy
93.8%
Sensitivity
98.3%
Specificity
Confusion matrix (5‑fold CV)
45
True positives
59
True negatives
1
False positives
3
False negatives
Performance from honest 5‑fold stratified cross‑validation on the curated benchmark.
Internal validation only.