Development of AOP-Informed Predictive Models for Screening Chemicals with Metabolic Dysfunction-Associated Steatotic Liver Disease (MASLD) Induction Potential

Abstract

Metabolic dysfunction-associated steatotic liver disease (MASLD) is a major liver disease, particularly influenced by drug or chemical interactions that induce hepatic steatosis. An adverse outcome pathway (AOP) can help explain these effects. A major target in this process is the peroxisome proliferator-activated receptor (PPAR), with three isoforms─PPAR-α, PPAR-β/δ, and PPAR-γ─each playing a crucial role in lipid metabolism. In this study, we developed predictive models to assess a chemical’s potential to induce MASLD using AOP information, particularly AOP-36, which describes the development of hepatic steatosis due to reduced activation of PPAR-α, PPAR-β/δ, and PPAR-γ. We collected data from the ChEMBL database for three PPAR targets and developed models for the ‘Standard Type’ IC50. After curation, the collected datasets include 456 data points for PPAR-α, 310 for PPAR-β/δ, and 1006 for PPAR-γ. Initially, we developed machine learning-based quantitative structure–activity relationship (QSAR) models for all the targets, which show moderate predictive performance (Q2F1/F2 ∼ 0.57–0.62). Then similarity-based read-across predictions were performed, where all datasets showed improved predictive performance (Q2F1/F2 ∼ 0.65–0.70). We also developed machine learning-based quantitative read-across structure–activity relationship (q-RASAR) models, which also show improved predictive performance (Q2F1/F2 ∼ 0.65–0.70).