risk, when combined into a polygenic score, they offer increasing power to capture information on individual patterns of disease predisposition with the potential to influence clinical management. • The generation of polygenic scores based on overall T2D predisposition can identify individuals with a high future risk of diabetes who may benefit from targeted interventions. • The generation of polygenic scores based on overall T1D risk can identify individuals who may benefit from early interventions to forestall the risk of T1D, and also supports the identification of those with lateronset diabetes who have an autoimmune etiology, for whom early recourse to insulin therapy may be advantageous. • The generation of partitioned polygenic scores which capture aspects of the etiological and clinical heterogeneity that contributes to variable clinical outcomes in those with T2D has potential to deliver clinical benefit through enhanced capacity to predict disease progression, complication risk, and response to pharmacological and behavioral interventions. • Polygenic scores have predominantly been derived from genetic studies performed in European populations and have suboptimal ability to capture risk in individuals of non-European origin. • Though there are a number of technical and logistical issues to be addressed before the clinical utility of polygenic scores can be fully enumerated, increasing utilization of polygenic scores within diabetes clinical practice is likely to be an important component of efforts to deliver precision medicine for those who have, or are at risk of, diabetes. 1. Introduction Diabetes is already one of the major contributors to death and ill-health globally, and its prevalence continues to rise. Current projections estimate almost 500M affected by diabetes as of 2017 (and almost 700M by 2045), most of this in the form of type 2 diabetes (T2D) [1]. Escalating rates of T2D speak to the limits of current strategies for prevention, whether they involve lifestyle interventions (for example through dietary modification and increased physical activity) or pharmacotherapy. At the same time, the burden of disease arising from the complications of inadequately controlled diabetes (manifest as renal failure, vision loss, amputation, and accelerated vascular disease) highlights the urgent need for major improvements regarding both the timely diagnosis of diabetes (since much damage is ADVANCE ARTICLE: Endocrine Reviews Downloaded from https://academic.oup.com/edrv/advance-article-abstract/doi/10.1210/er.2019-00088/5535575 by 81225740 user on 24 July 2019 ADVANCE ARTICLE Endocrine Reviews; Copyright 2019 DOI: 10.1210/er.2019-00088 3 initiated whilst the disease is subclinical) and the management of those with established disease. The condition that we currently label as “type 2 diabetes” represents a convenient, but likely suboptimal, construct for the application of 21st century medicine. Though individuals with established T2D have a generalized metabolic derangement (typically associated with hyperlipidemia, adiposity, disturbed hepatic metabolism and the like), formal diagnosis rests entirely upon a single metabolic component (glucose), itself the end-result of multiple metabolic processes. The diagnosis of diabetes depends on numeric thresholds placed within continuous distributions of (fasting, random or postprandial) glucose and/or glycated hemoglobin levels. These thresholds were initially based around the observed relationships between levels of hyperglycemia and the incidence of specific diabetic complications, such as retinopathy, but they may not be equally discriminating for the macrovascular complications [2]. Crucially, T2D remains effectively a diagnosis of exclusion, made after those with hyperglycemia attributable to more defined causation including islet autoimmunity (type 1 diabetes [T1D]), highly penetrant genetic effects (e.g. maturity onset diabetes of the young [MODY]) and certain specified exposures (steroids, pancreatitis, pregnancy) have been excluded. Those left with the diagnosis of T2D demonstrate considerable heterogeneity with respect to presentation, clinical course, and response to available therapies, yet clinical pathways tend to be based around universally-applied algorithms that take little, if any, account of that heterogeneity [3-5]. Human genetics provides a powerful set of approaches for addressing some of these challenges, delivering both an improved understanding of the mechanisms contributing to the development of diabetes, and opportunities for direct translational benefit [6]. Both common major subtypes of diabetes (T1D, T2D) are complex, multifactorial traits: that is, an individual’s risk of developing either of these conditions is influenced by the combination of genetic variation at multiple sites across the genome, acting in concert with factors within the external (e.g. nutritional availability, socio-economic status) and internal (e.g. microbiome, metabolic memory) environment [7,8]. Over the past decade, large-scale genetic studies (typically in the form of genome-wide association studies [GWAS]) have identified over 400 distinct genetic signals influencing T2D risk [9] and over 50 with impact on T1D predisposition [10]. Most of these DNA sequence variants are widely shared within and between populations, in contrast