Endocrinology in the multi-omics era

Over the past decade, technological advances have enabled cost-efficient, high-throughput analysis of different types of omics data in large human cohorts. Here, we explore insights into the pathophysiology of metabolic disorders revealed through multi-omics studies, discuss novel computational analysis techniques and look at the field’s future directions.

Key advances

• In-depth characterization of glycaemic patterns obtained from continuous glucose monitoring (CGM) devices from thousands of individuals without diabetes mellitus provided a map of reference values for CGM-derived measures and an invaluable tool for future research on glucose control, glycaemic variability and associations with clinical measures representing various aspects of human health².

• The creation of an atlas of blood analytes associated with variations in BMI, based on multi-omics profiling, demonstrated metabolic heterogeneity within similar BMI classes as well as trajectories of different omics measures in response to lifestyle intervention³.

• Single-cell analysis identified two transcriptionally and functionally distinct β-cell subtypes that undergo an abundance shift during type 2 diabetes mellitus (T2DM) progression; gene regulatory network analyses were associated with the genetic risk of T2DM, implying a potential causal role of β-cell subtype identity in T2DM development⁴.

• Multi-omics profiling unravelled gut bacteria associated with insulin resistance that manifest a distinct pattern of carbohydrate metabolism; insulin-sensitivity-associated bacteria ameliorated host phenotypes in a mouse model, suggesting a novel mechanism by which gut microbial carbohydrate metabolism contributes to insulin resistance⁵.

• The development of foundation models (self-supervised learning models, such as RETFound⁹) has emerged as an important advancement in data analysis methods, enabling models to perform a wide range of tasks with minimal labelled data and to adapt to various types of medical data.