PROJECT SUMMARY/ABSTRACT Millions of Americans are at risk of type 2 diabetes, but we lack a precise understanding of mechanisms causing high blood glucose levels, life-altering complications, and premature death. Current criteria for type 2 diabetes use glucose and HbA1c diagnostic cutoffs based on a link to retinopathy and thus do not account for risk of other microvascular and macrovascular complications. To improve approaches to defining prediabetes and diabetes, risk stratification is critical. Recent studies from our group and others demonstrate the potential to classify patients with early diabetes that may ultimately result in personalized treatment and better outcomes. Despite the reproducibility within existing subtype definitions, including genetically determined clusters and new-onset diabetes clusters, there is no unifying theory or strategy for type 2 diabetes classification, which must integrate different data types. Type 2 diabetes needs to be re-defined so that precise and accurate diagnosis can be provided to patients, and the appropriate mechanism-guided therapies applied, to alleviate suffering and prevent complications. We plan to challenge the current paradigm that defines diabetes and prediabetes based solely on glucose and HbA1c. Our overarching goal is to integrate clinical and multi-omic data into prediabetes and type 2 diabetes classification and build prediction models of disease trajectories over the lifespan. Our central hypothesis is that multi-omic data can refine definitions of type 2 diabetes, identify those patients most like to develop complications, and better identify people at risk for future development of type 2 diabetes. The following three specific aims are proposed: Aim 1: We will use multi-omic data to refine existing prediabetes and type 2 diabetes definitions and define new distinct subtypes of these conditions. Aim 2: We will describe the associations of diabetes subtypes with microvascular and macrovascular complications and identify environmental contexts that modify disease trajectories. Aim 3: We will determine whether subtypes based on high dimensional clinical and multi-omic data in MESA, ARIC, HCHS/SOL, All of Us and other cohorts from the consortium can be recapitulated using clinical biomarkers. Our expected outcomes are (1) characterization of more homogenous subgroups of diabetes patients among US populations of multi-ethnic groups, (2) identification of lifestyle and/or medical interventions that reduce type 2 diabetes risk or risk of future complications for any given subgroup, and (3) new diagnostic tools that can be translated to a clinical setting to more accurately classify type 2 diabetes.