["The Journal of Physiology, EarlyView. ", "\nAbstract figure legend Machine learning analysis of circulating microRNA (miRNA) profiles identified a minimal set of biomarkers that distinguish individuals with obesity from lean individuals both before and after weight‐loss intervention. Comparative analyses revealed heterogeneous molecular responses to weight reduction, with some miRNAs showing partial normalization while others remained persistently dysregulated. Integration of miRNA signatures with target gene interaction networks and pathway enrichment analyses indicated that persistent signals are linked to immune–metabolic regulatory pathways and systemic metabolic control. These findings highlight the presence of both reversible and core molecular components underlying obesity and suggest that circulating miRNAs can serve as informative biomarkers of obesity‐related dysregulation and treatment response. Overall, this work demonstrates the utility of machine learning for uncovering biologically meaningful molecular signatures and provides new insight into the complex regulatory landscape of obesity.\n\n\n\n\n\n\n\n\n\nAbstract\nObesity is a complex metabolic disease characterized by systemic metabolic and inflammatory dysregulation, yet the molecular signatures underlying these processes remain incompletely understood. Circulating microRNAs (miRNAs) have emerged as promising biomarkers capable of capturing systemic regulatory changes associated with obesity. In this study, we investigated whether machine learning (ML) could identify obesity‐discriminative circulating miRNA signatures and assess their persistence following weight loss. Circulating miRNA profiles from lean individuals and individuals with obesity before and after a weight‐loss intervention were analysed using ML‐based classification frameworks combined with feature selection and multiple classifier models. Comparative analyses of miRNA signatures were further integrated with target gene interaction networks and pathway enrichment analyses to explore the biological processes associated with obesity and weight‐loss responses. The ML models identified a small set of circulating miRNAs capable of distinguishing individuals with obesity from lean individuals both before and after weight loss. Comparative analyses revealed that some miRNAs showed partial normalization after weight reduction, whereas others remained persistently dysregulated. Network and pathway analyses suggested that persistent miRNA signals are linked to regulatory processes involved in immune–metabolic interactions and systemic metabolic control. These findings indicate that circulating miRNAs capture both reversible and persistent molecular components of obesity and may serve as informative biomarkers of obesity‐related dysregulation. Overall, this work demonstrates the utility of ML for uncovering biologically meaningful miRNA signatures and provides new insight into the molecular complexity of obesity and its response to weight‐loss interventions.\n\n\n\n\n\n\n\n\n\nKey points\n\nMachine learning (ML) identified a minimal circulating microRNA (miRNA) signature that robustly discriminates obesity (baseline and following weight loss) from lean status, with performance comparable to transcriptomic models.\nSeveral miRNAs remained persistently dysregulated after weight loss, suggesting core obesity‐related pathways and potential predisposition to weight regain.\nOther miRNAs normalized following weight loss, indicating reversible, metabolically responsive mechanisms (e.g. glucose regulation).\n\n\n"]