Machine learning-guided miRNA screening and framework nucleic acid-based nanotherapy for obesity-induced comorbidities

Obesity is a global health crisis driving insulin resistance (IR), type 2 diabetes (T2D), and metabolic dysfunction-associated steatotic liver disease (MASLD). Chronic low-grade inflammation in adipose tissue, mainly driven by macrophages, lies at the core of this process. Current treatments often lose effectiveness, cause weight regain, or produce systemic side effects. New strategies with better targeting, higher efficacy, and lower toxicity are urgently needed. MicroRNA (miRNA) therapy holds great promise but faces two major hurdles, namely selecting the right miRNA and the lack of an efficient delivery system. In this study, we used a multi-method machine learning ensemble combining PCA, random forest regression, and expert-weighted scoring across nine features to prioritize obesity-related miRNAs, and miR-146a-5p emerged as the top candidate. Guided by this screening, we engineered a sticky-end tetrahedral framework nucleic acid (stFNA) loaded with miRNA via vertex-complementary base pairing. The resulting miR-stFNA nanocomplex has a uniform size, good stability, efficient macrophage uptake, and lysosomal escape. In vitro, miR-stFNA suppressed reactive oxygen species (ROS), nitric oxide (NO), and pro-inflammatory cytokines by curbing MAPK, NF-κB, and PI3K pathways in activated macrophages, outperforming each component alone. In the obese mouse model, miR-stFNA reduced weight gain, improved insulin sensitivity, and decreased liver fat accumulation. This work integrates machine learning-based miRNA selection with DNA origami delivery, providing a smart nanotherapy for obesity-related complications.