Scientists have developed a computational framework, DeepDDI, that accurately predicts and generates 86 types of drug-drug and drug-food interactions as outputs of human-readable sentences, which allows in-depth understanding of the drug-drug and drug-food interactions.