Multi-tissue drug toxicity prediction with mechanism decomposition.
ToxiPred returns per-tissue cellular Drug Sensitivity Scores (DSS) and decomposes each prediction into pathway-level Hadamard contributions. Novel chemistry is handled by a five-seed neural ensemble with a Tanimoto similarity indicator.
A pathway-level read-out for every query.
Each prediction returns per-tissue DSS across 24 lineages, the top-affected tissues, and a ranked mechanism panel - apoptosis, OXPHOS, cholesterol biosynthesis, inflammatory response, and ~1,400 other pathways scored from the bottleneck ensemble.
- Apoptotic - BH3-only proteins, p53, intrinsic cell death
- Lipid / Sterol - cholesterol biosynthesis, fatty-acid metabolism
- Stress / UPR - protein secretion, ROS, inflammatory response
- Cell-cycle / DNA-repair - G2M, MYC targets, ATAD5, DSB-repair
Sertraline
tetralin SSRIPipeline
-
01
Resolve
Canonical SMILES match against the indexed compound atlas.
-
02
Predict
Exact match returns measured DSS. Novel chemistry passes through the five-seed bottleneck ensemble.
-
03
Decompose
Hadamard pathway contributions are scored and gated by the toxicity band (DSSmean > 5.0).
-
04
Report
Per-tissue DSS, top-3 mechanisms, dominant family.