Integrated meta-analysis of public human pancreatic single-cell transcriptomes reveals distinct beta-cell state trajectories across aging and diabetes

Abstract

Human pancreas single-cell RNA sequencing (scRNA-seq) studies have revealed extensive islet heterogeneity, yet cross-study integration remains limited by cohort-and platform-specific effects. Here, we assembled a unified atlas of >266,000 human pancreatic cells by harmonizing 18 publicly available scRNA-seq datasets spanning diverse technologies and donor phenotypes. Trajectory-based analyses resolved three beta-cell state trajectories associated with distinct stress axes. One trajectory reflects aging-associated transcriptional drift with progressive ER stress activation. A second captures diabetes-associated remodeling characterized by combined ER stress and induction of exocrine-like gene programs. A third highlights a metabolic-stress-associated program linked to lipid metabolism and polyhormonal transcriptional signatures in non-diabetic donors with elevated metabolic burden. In contrast to the relative stability of alpha-cell states, beta-cell identity programs eroded along specific trajectories, often preceding marked reductions in INS expression. Together, this integrated resource provides a scalable framework for dissecting human beta-cell plasticity and dysfunction using public single-cell transcriptomic data.