ORF15-05 - A Systems-Level Map of the Circulating Serum Proteome Across the Fibrosis and Inflammation Spectrum in Biopsy-Proven Metabolic Dysfunction-Associated Steatohepatitis

Metabolic dysfunction-associated steatohepatitis (MASH) is increasingly encountered across hepatology, bariatric, and metabolic care, yet molecular tools that capture disease biology and accurately identify clinically meaningful fibrosis remain limited. We profiled 165 well-phenotyped adults with biopsy-confirmed MASH (73 men, 92 women; 119 with obesity) recruited from two tertiary metabolic and gastroenterology centers. Serum proteomics were measured using the SomaScan 7K platform (SomaLogic Inc), quantifying 6,461 shared human proteins. Proteomic trajectories across NAFLD Activity Score (NAS) components and fibrosis stages were interrogated using heatmaps, correlation analyses, and ANOVA/ANCOVA models adjusted for age, sex, BMI, and HOMA-IR, complemented by weighted gene co-expression network analysis (WGCNA) and pathway enrichment. Across the MASH spectrum, 2,046 proteins associated with fibrosis severity (FDR < 0.05) and 2,263 with NAS, with additional signatures linked to hypertension, diabetes, dyslipidemia, and obesity. Principal component analyses distinguished metabolic versus hepatology cohorts while also revealing progressive proteomic convergence with increasing NAS and fibrosis, consistent with a shared severity program across care settings. Fibrosis progression was characterized by increased mTORC1/MYC signaling and mitochondrial stress programs and decreased xenobiotic metabolism and angiogenesis. WGCNA identified hepatotoxic modules strongly associated (FDR < 10^-5) with steatosis (r=0.46), lobular inflammation (r=0.53), hepatocellular ballooning (r=0.38), transaminases (r=0.35-0.40), and circulating hormones including GDF-15 (r=0.37 total; r=0.31 intact) and adiponectin (r=-0.33). As MASH severity increased, pathway enrichment (deCODE) indicated corresponding increases in neuro-immune and stress-related pathways, including signatures related to depression, mood, pain disorders, and kidney disease, whereas HDL-related coronary artery disease signatures were inversely enriched. We then applied a stratified cross-validated 70:30 train-test split with cross-validated model tuning confined to the training set, and built a gradient-boosting classifier combining five clinical variables with four newly discovered proteomic features (not previously used in MASH fibrosis models). This model could identify at-risk MASH with F2-F3 fibrosis in the independent test set (AUC=0.87; accuracy=0.88; sensitivity=0.99; specificity=0.83), outperforming conventional biomarker-, imaging-, and algorithm-based non-invasive tests. Together, this first-in-kind comprehensive SomaScan 7K investigation in biopsy-proven MASH, focused on clinically meaningful F2-F3 disease, provides a systems-level molecular map and a novel diagnostic framework that is currently being evaluated in larger independent cohorts.