Abstract
<title>Abstract</title> <p>Background Metabolic-associated steatotic liver disease (MASLD) is a heterogeneous disease with patients displaying disparate prognosis or response to candidate therapies. We hypothesized that subgroups of patients may have different intrahepatic molecular presentation and could benefit from targeted treatments. Methods We analysed anonymized publicly available data, applying a subtype of unsupervised neural network algorithm, called self-organizing maps. The algorithm learnt hepatic gene expression pattern in the discovery cohort GSE135251 and categorized patients into molecular subtypes. Independent cohorts including Caucasian (GSE130970, GSE130991, E-MTAB-9815, GSE225740), Japanese (GSE167523, GSE193066) and Hispanic (GSE185051) donors were used for validation. Differential expression, pathway analysis and cell type deconvolution were performed with state-of-the art methods. Results We identified five molecular subtypes (s1 to s5). s1 was observed in n = 841 (79.4%) of donors in 8 independent cohorts. s5 was observed in 5 cohorts (n = 78, 7.4% of donors). s4 was observed in 3 cohorts (n = 83, 8.8% of donors). Each subtype had similar pathway dysregulation in both mild and advanced fibrosis. s1, s5 and s4 were characterized by increased cellular stress and injury, lipid and glucose metabolism and fibrosis and decreased mitochondrial function. s4 had the highest expression of cancer-associated pathways. s1 and s5 showed milder pathway dysregulation than s4, especially less inflammation pathways. s1 was depleted of endothelial cells. s2 and s3 had an atypical presentation with lower expression of many inflammation and cancer-associated pathways than non-MASLD controls. s2 (n = 13) and s3 (n = 34) were observed only in the discovery cohort and had low expression and activity of THRB. Conclusions Across independent cohorts, liver transcriptomics defined five histology‑independent MASLD subtypes with distinct pathway and immune-stromal programmes, supporting disease segmentation with implications for prognosis and therapeutic response.</p>