Today, we’re honoring the lives of the 32 Hokies who were tragically taken from their loved ones and the Virginia Tech community on April 16, 2007. On the 2025 Day of Remembrance, we're honoring the lasting impact they continue to have on our world. #VTWeRemember 🧡

Our article is now published in @NatureMedicine! https://t.co/3KdeyL7vTc @miriam_udler @broadinstitute @CGM_MGH

U-M BME researchers, led by Associate Professor Sriram Chandrasekaran and postdoctoral fellow Awanti Sambarey, developed a multimodal AI model to predict treatment outcomes of TB patients, which may one day lead to the personalization of TB treatment.

Special thanks to Aaron Deutsch, @miriam_udler @josep_mercader @JoseFlorezMDPhD @kjgaulton @kewesterman @AlisaManningPhD @AC_Williamson @l_szczerbinski @MelinaClaussnit & our other authors for their contributions, as well as @broadinstitute @MGH_RI @CGM_MGH for their support

5/6) The Lipodystrophy clusters (A) contributed a larger proportion of polygenic T2D risk in East Asians and (B) also explained a large difference in BMI-adjusted T2D risk between Europeans and East Asians.

4/6) The 12 genetic clusters were distinctly enriched for single-cell regulatory regions. Insulin-deficient clusters, such as Beta Cell 1 and 2, were enriched for epigenetic annotations in pancreatic islets, while Lipodystrophy clusters achieved significance in adipose nuclei.

3/6) We used individual- and GWAS-level data to characterize the clusters. Interesting findings: the new Lipodystrophy 2 cluster was associated with increased NAFLD risk, while the new Cholesterol cluster was associated with decreased CAD risk.

2/6) We performed soft clustering on 650 T2D-associated, multi-ancestry genetic variants and 110 related traits. The resulting clusters included three novel clusters related to lipodystrophy, cholesterol and bilirubin.

1/6) Type 2 diabetes (T2D) is a heterogeneous disease for which causal pathways are incompletely understood. In our new preprint, we identify 12 genetic T2D subtypes and demonstrate how they may explain variation in polygenic risk across populations