Graph-Based Machine Learning Models Improve Prediction of Drug-Induced Liver Injury and Cellular Toxicity

A team of researchers published a preprint on bioRxiv describing a comprehensive quantitative structure-activity relationship (QSAR) modeling pipeline designed to predict in vitro and in vivo chemical toxicity. The pipeline integrates data preprocessing, molecular fingerprints, molecular graph representations, and a range of machine learning models—including classical approaches, graph neural networks (GNNs), and graph transformers (GTs)—benchmarked against PubChem assay data for caspase-3/7 activation and mitochondrial membrane potential (MMP) disruption. For FDA Drug-Induced Liver Injury (DILI) prediction, the full consensus model achieved an AUC of 0.69 and Graphormer achieved an F1 score of 0.79, both substantially surpassing the previous best reported AUC of 0.63 and F1 of 0.65. The study found that graph-based models excelled when active compound counts were large, while classical models and GTs performed better on highly imbalanced datasets with few active compounds. Mechanistic analysis identified phenolic compounds with a para-hydroxyphenyl motif and lipophilic long-chain alkyl compounds as capable of collapsing mitochondrial membrane potential and activating caspases-3/7. Cell-line-specific structural motifs were also identified, including 1,1-dichloroethane and chlorobenzene for HEK293 cells, an epoxide fragment for SK-N-SH neuroblastoma cells, and tetramethylcyclohexene and acetaldehyde fragments for H-4-II-E rat hepatoma cells. The authors plan to incorporate biological activity data, toxicity literature, large language models, and agentic AI to further refine the framework.