ML-powered ADMET predictions, explainable at the atom level. Catch toxicity signals early — before they cost you a clinical trial.
Predict absorption, distribution, metabolism, excretion and toxicity from a SMILES string in under 1.5 seconds — with atom-level explanations so you know why, not just what.
Binary classifier trained on TDC canonical splits (AUROC >0.84). Flags compounds that risk QT prolongation before they reach wet-lab assays.
Gate 1 · Q1 2027Drug-induced liver injury and genotoxicity screening. Same API endpoint, same sub-second latency, same atom-level explainability.
Gate 1 · Q1 2027GradientExplainer identifies exactly which atoms drive each prediction, rendered as a 2D SVG heat map on the molecular structure.
Gate 1 · Q1 2027ONNX-optimised inference on Cloud Run. Single molecule in <1.5 s, batch of 1,000 in under a minute.
SHAP GradientExplainer shows which atoms drive each prediction. Defensible to medicinal chemists and regulatory reviewers.
REST API with OpenAPI spec, JWT + API key auth, async batch jobs and audit logs. Built for enterprise IT security review.
Chemprop v2 DMPNN — consistently top-ranked on TDC ADMET benchmarks. Not a wrapper around a public API.
Every prediction traces back to training dataset, model version and AUROC. Complete reproducibility for regulatory dossiers.
Built by someone who has worked inside Genentech, Novartis, Celgene and Novo Nordisk — not just read the papers.
ADMET predictions for your pipeline, Q1 2027. First 50 sign-ups get a free 90-day trial at launch.
✓ You're on the list. We'll reach out before launch.