A Protein Interactome-Based Framework Reveals the General Toxicity of Chemicals

Humans are exposed to various chemicals, and understanding their toxicity is essential for safeguarding their health. However, current toxicological methods often focus on specific diseases or chemicals, missing broader patterns of toxicity. Here, we propose a general framework that conceptualizes toxicity as arising from protein-interaction-mediated relationships between chemicals and diseases. Using the Comparative Toxicogenomics Database, we show that the network proximity between chemical targets and disease-associated proteins in the protein interactome predicts chemical-disease associations. We validate this finding through zebrafish experiments for acute toxicity and analyses of human exposome data sets for chronic chemical-disease correlation. We highlight the applications of our framework in predicting a previously unknown toxic effect of a fungicide, explaining drug cytotoxicity in a COVID drug screening, and uncovering the overlooked ″indirect-hit″ chemical-disease toxicity pattern. Our approach establishes a new paradigm for toxicity research, offering insights into the health risks posed by chemicals and addressing public health challenges.