Cheminformatics-based read-across prioritization of food additives with obesogenic potential and PPARγ agonist activity.

Food additives are widely used in processed foods, yet many remain insufficiently evaluated for potential metabolic-disrupting effects. In this study, we developed a cheminformatics-based read-across approach integrating curated food additive inventories with Tox21 high-throughput screening (HTS) data to prioritize food additives with potential obesogenic activity. We integrated 7917 compounds from the Tox21 dataset with 3185 compounds from curated food additive inventories, followed by rigorous data curation to yield a dataset comprising 10,077 compounds. Dimensionality reduction and clustering analyses identified 428 structurally defined clusters. Enrichment analysis of Tox21 assay endpoints identified five clusters significantly associated with peroxisome proliferator-activated receptor γ (PPARγ) agonist activity. From these PPARγ activity-enriched clusters, eight food additives not previously evaluated in Tox21 were prioritized for experimental validation and assessed in an adipogenic differentiation model of human bone marrow-derived mesenchymal stem cells (hBM-MSCs). Among them, benzyl disulfide (BD), hesperetin (HSP), and syringaldehyde (SD) significantly enhanced adiponectin biosynthesis and lipid accumulation. Pharmacological profiling further demonstrated that BD and HSP directly bound to PPARγ and induced coactivator recruitment. Collectively, these findings support the use of integrated cheminformatics read-across analysis and targeted in vitro testing for hazard identification and prioritization of food additives with potential obesogenic activity. The results provide a basis for further mechanistic and exposure-informed assessment of the prioritized candidates.