CardioRisk-DB
A multi-omics knowledgebase for phenotype-anchored cardiotoxicity risk prediction
Integrating transcriptomics, genetic evidence, and causal inference to identify cardiotoxic drugs and molecular mechanisms. Powered by PAMD and Split-GSVA contrastive modeling.
Database Overview
Scientific Framework
Drug-induced transcriptional responses often contain non-specific cellular stress signals (e.g., heat shock, DNA repair) that obscure true cardiotoxic mechanisms.
Our framework utilizes Phenotype-Anchored Multi-omics Distillation (PAMD) combined with contrastive Split-GSVA scoring to filter out generic background noise and isolate cardiotoxicity-specific molecular signatures.
Explore Modules
Signature Architecture
Distribution of PAMD selected genes across the Upregulated (Risk) and Downregulated (Protective) directions.
Leave-One-Drug-Out (LODO) Audit & Stability
Model stability when selectively removing specific chemical classes from the training manifold.
Causal Forest Plot: Gene-Disease Associations
Mendelian Randomization (MR) evidence linking signature genes to clinical cardiovascular outcomes. Error bars indicate 95% CI.
Upload Data
Prediction Report
About CardioRisk-DB
CardioRisk-DB is an open-source, multi-omics knowledgebase dedicated to evaluating the cardiotoxicity of environmental chemicals and pharmaceutical drugs.
Citation
If you use CardioRisk-DB in your research, please cite:
Zhang J, et al. (2026). Phenotype-Anchored Molecular Distillation reveals contrastive transcriptomic signatures for robust cardiotoxicity prediction.