Artificial intelligence in cardio-oncology: decoding mechanisms, predicting toxicity, and personalizing cancer therapy

Figures

• 

  Figure 1. AI-Mediated workflow from multi-omics to clinical phenotypes.

• 

  Figure 2. Core mechanism framework for AI recognition of heart protective drugs. In the context of drug repurposing, AI algorithms specifically identify cardioprotective agents through several key mechanisms: (1) feature selection methods (e.g., LASSO regularization, random forest importance scores) prioritize genes or proteins most strongly associated with both disease pathology and potential drug interactions; (2) network-based approaches map these prioritized targets onto protein-protein interaction networks and pathway databases to identify druggable nodes within disease modules; (3) molecular docking and dynamics simulations then computationally screen compound libraries against these prioritized targets to predict binding affinities and stability, with binding energies ≤−5 kcal/mol typically indicating stable interactions; (4) integration of toxicogenomic databases (e.g., CTD, DisGeNET) enables early filtering of compounds with potential adverse effects; (5) machine learning classifiers trained on known drug-cardiotoxicity associations can predict the likelihood of both therapeutic efficacy and cardiac safety for candidate compounds. These computational pipelines systematically…