Tissues develop and function within a highly complex microenvironment, where diverse cell types interact in tightly regulated spatial and temporal patterns. Through distinct, relay-like waves of activity, these cells collectively shape tissue formation, ensuring that each component emerges in the right place at the right time. Such coordinated cues establish the structural and biochemical framework that drives cell differentiation and tissue organization. Accurately modelling this process in humans requires the development of complex multicellular systems. Here, we pair spatial transcriptomics of the human foetal pancreas with an induced pluripotent stem cell (iPSC)-based multilineage model that faithfully recapitulate the complex, hierarchical processes underlying human pancreatic islet formation. We show that iPSC-derived pancreatic mesodermal lineages direct endocrine commitment from pancreatic progenitors, by suppressing off-target fates and orchestrating niche-mediated spatio-temporal cues that promote beta-cell differentiation with enhanced insulin secretion. Our results identify a vascular-rich niche, featuring pancreatic pericytes, which activates a neural repulsion program shaping the islet microenvironment. We benchmark our in vitro multilineage organoid system against spatial transcriptomics of the human foetal pancreas. Together, these findings identify the key cellular actors and contact-dependent mechanisms that build the human endocrine pancreas, providing a critical model for studying human islet development and disease.
Magnification-Aware Distillation (MAD): A Self-Supervised Framework for Unified Representation Learning in Gigapixel Whole-Slide Images
arXiv:2512.14796v1 Announce Type: cross Abstract: Whole-slide images (WSIs) contain tissue information distributed across multiple magnification levels, yet most self-supervised methods treat these scales as independent


