Machine Learning Helps ID Molecular Mechanisms of Pancreatic Islet Beta Cell Subtypes in Type 2 Diabetes

In a new study appearing in Nature Genetics, researchers identify two distinct pancreatic islet beta cells. These cells are known to be dysfunctional in patients with type 2 diabetes (T2D), however, researchers didn't have a clear understanding of the underlying mechanisms that cause these cells to go awry, the authors note. By using machine learning approaches, the researchers studied measurements of chromatin accessibility, gene expression, and function in single beta cells from 34 non-diabetic, pre-T2D, and T2D people, and identified two transcriptionally and functionally distinct beta cell subtypes of these cells. They also found that these cells undergo an "abundance shift" as the disease progresses. "The machine learning approach overcomes limitations of unsupervised dimensionality reduction methods for identifying disease-associated patterns in single-cell data from tissues of heterogenous donors and could have broad applications for interpreting single-cell maps of human tissues," the authors write. By analyzing genetic data from all the participants, the authors were also able to pinpoint that reduced HNF4A and HNF1A activity could be the reason behind T2D pathogenesis and is likely responsible for the shift in beta cell subtype identities as the disease progresses, the authors note.