Abstract
Background: Autism spectrum disorder (ASD) is increasingly linked to systemic metabolic dysfunction, potentially influenced by gut-brain axis dysregulation, but the underlying mechanisms remain unclear. Methods: We developed Personalized Metabolic Margin Mapping (PM(3)), a computational systems biology framework, to analyze RNA-seq data from 12 ASD and 12 control postmortem brain samples. The model focused on 158 curated metabolic genes selected for their roles in redox balance, mitochondrial function, neurodevelopment, and gut-brain interactions. Results: Using unsupervised machine learning (Isolation Forest) to detect outlier expression patterns, Euclidean distance, and percent expression difference metrics, PM(3) revealed a consistent downregulation of glycolysis (e.g., -5.4% in PFKM) and mitochondrial enzymes (e.g., -12% in SUCLA2). By incorporating cofactor dependency and subcellular localization, PM(3) identified a coordinated suppression of multivitamin transporters (e.g., -4.5% in SLC5A6, -3.5% in SLC19A2), potentially limiting cofactor availability and compounding energy deficits in ASD brains. Conclusions: These findings suggest a convergent metabolic dysregulation signature in ASD; wherein the subtle suppression of cofactor-dependent pathways may impair energy metabolism and neurodevelopment. We propose that chronic microbial lipopolysaccharide (LPS) exposure in ASD suppresses vitamin transporter function, initiating mitochondrial dysfunction and transcriptomic reprogramming. Validation in LPS-exposed systems using integrated transcriptomic-metabolomic analysis is warranted.