1331-OR: Hepatic IRS1 as a Therapeutic Target in Metabolic Dysfunction–Associated Liver Disease (MASLD): AI/ML-Driven Genetic Discovery and In Vivo Therapeutic Validation

Introduction and Objective: MASLD is a common polygenic condition, but current GWASs have only identified 14 loci due to a lack of scalable, clinically relevant phenotypes. Methods: Using machine learning on four UKBB data modalities (MRI, DXA, plasma metabolites, blood biomarkers; n=25,474 to 262,927), we predicted liver fat (LF) and adipose tissue volume (ATV), increasing statistical power. A GWAS of LF, adjusting for ATV, identified liver-specific genes. Liver-targeted siRNAs validated key targets in vitro and in vivo in mice and monkeys. Results: GWAS identified 480 loci linked to LF accumulation, showing strong genetic correlation with predicted LF (R2 = 0.79-0.97), and MASLD (R2 = 0.73-0.87). Novel de novo lipogenesis (DNL) genes were found, with IRS1 being the strongest association (p = 2 x 10-82) and supported across all four modalities. A potent, selective liver-targeted siRNA against IRS1 (IC50~0.04 nM) suppressed lipid accumulation in human hepatocytes. In DIO mice a single dose (0.3-10 mg/kg) reduced hepatic IRS1 expression (85%), DNL (59%) and triglycerides (45%) after one month, significantly improving NAS score (2.2 ± 0.2 vs. vehicle 4.3 ± 0.2, p = 0.0002). Cynomolgus monkeys (3 and 10 mg/kg) showed sustained hepatic IRS1 reduction for 28 days without affecting glucose homeostasis. Conclusion: Liver-targeted siRNA effectively targets IRS1, showing efficacy and tolerability in preclinical MASLD models. Disclosure S. Satapati: Stock/Shareholder; Current; insitro. D. Lloyd: None.