Computational Analysis and Prediction of CYP1A2-Related Toxicants for Safer Drug Discovery

Drug metabolism, primarily driven by cytochrome P450 (CYP450) enzymes, plays a vital role in drug efficacy and safety by transforming xenobiotics into active or detoxified forms. Among these enzymes, CYP1A2 is responsible for approximately 10\% of drug metabolism and is associated with toxicological concerns due to its role in biotransformation and mechanism-based inhibition. To identify potential CYP1A2-related toxicants, we curated a reference set of 10 tox- icologically relevant substrates and 10 inhibitors, then conducted molecular simi- larity searches using RDKit against the drug-like subset of the ZINC database. We applied structure-based molecular docking with Schrödinger’s Glide to evaluate binding interactions and affinities, enabling refinement of the datasets. K-means clustering with Elbow method analysis guided the selection of optimal similar- ity thresholds, yielding two expanded datasets: 2,973 substrate-like compounds (average similarity 0.63) and 2,433 inhibitor-like compounds (average similarity 0.53). These datasets revealed structural features associated with known toxicants such as phenacetin, 2-acetylaminofluorene, and myristicin. The resulting com- pound collections provide a robust foundation for training predictive models of CYP1A2-related toxicity and support safer drug discovery through the early identi- fication of compounds with high metabolic liability. The code and created datasets are available at https://github.com/yaow1004/CYP1A2_Toxicants_Dataset.