Plasticizer-responsive molecular axes in heart failure: a subtype-aware toxicogenomic framework relevant to environmental health.

Environmental exposure to synthetic chemicals has been associated with cardiovascular risk, yet mechanistic connections between real-world chemical mixtures and heart failure heterogeneity remain incompletely characterized. Here we present a mixture-aware toxicogenomic framework that integrates multi-cohort human cardiac transcriptomes with function-based grouping of 37 plasticizers. Using multi-resource, consensus-derived target profiles, we organized plasticizers into six functional clusters, enabling classification beyond chemical structure. Rather than relying on curated disease gene sets or single-cohort differential expression thresholds, we used a cross-validated, one-vs-rest machine learning strategy to derive subtype-resolved heart failure (HF) signatures directly from multi-cohort transcriptomes and used these reproducible signatures as an anchor for exposure-oriented mechanistic interpretation. We prioritized cross-cohort generalizability and signature stability by focusing on subtype-discriminative genes that recur consistently across validation iterations. Integrative analyses suggested associations between plasticizer clusters and HF-relevant pathogenic axes, including inflammatory remodeling, metabolic stress, and extracellular matrix processes, with subtype-resolved enrichment patterns consistent with VEGF/MAPK and hormonal signaling in dilated cardiomyopathy (DCM) and TNF/IL-17 and the diabetes-associated AGE-RAGE axis in ischemic cardiomyopathy (ICM). Shared signals included matrix remodeling and MMP-related programs across both subtypes. Overall, this scalable framework provides interpretable exposure-linked disease axes to support hypothesis generation in environmental cardiotoxicity and to inform consideration of chemical mixtures in cardiovascular risk assessment and monitoring.