Advancing Computational Toxicology by Interpretable Machine Learning

Chemical toxicityevaluations for drugs, consumer products, andenvironmental chemicals have a critical impact on human health. Traditionalanimal models to evaluate chemical toxicity are expensive, time-consuming,and often fail to detect toxicants in humans. Computational toxicologyis a promising alternative approach that utilizes machine learning(ML) and deep learning (DL) techniques to predict the toxicity potentialsof chemicals. Although the applications of ML- and DL-based computationalmodels in chemical toxicity predictions are attractive, many toxicitymodels are "black boxes" in nature and difficult tointerpret by toxicologists, which hampers the chemical risk assessmentsusing these models. The recent progress of interpretable ML (IML)in the computer science field meets this urgent need to unveil theunderlying toxicity mechanisms and elucidate the domain knowledgeof toxicity models. In this review, we focused on the applicationsof IML in computational toxicology, including toxicity feature data,model interpretation methods, use of knowledge base frameworks inIML development, and recent applications. The challenges and futuredirections of IML modeling in toxicology are also discussed. We hopethis review can encourage efforts in developing interpretable modelswith new IML algorithms that can assist new chemical assessments byillustrating toxicity mechanisms in humans.