Exploration of Genetically Relevant Representations of Glucose Tolerance Tests Among Mexican Americans in Starr County, Texas

As obesity rates have risen, the prevalence of concomitant metabolic disorders has increased dramatically. Globally, around 415 million people have diabetes (9% of adults). Over the next decade, the number of affected cases is estimated to increase to 642 million people, making type 2 diabetes one of the fastest growing global epidemics. Rising prevalence, the severity of comorbidities, and relatively high estimates of heritability have lead type 2 diabetes to be the focus of decades of genetic research. Despite identifying \u3e80 loci associated with type 2 diabetes risk only ∼20% of its heritability is explained [1]. glucose dysregulation traits that are genetically enriched, may lead to the identification of newer loci associated with type 2 diabetes. ^ We propose a novel application of canonical correlation analysis as a way to extract genetically enriched traits. Canonical correlation analysis identifies the best function of genetic variation that is maximally correlated with the function of phenotypic variation. Via simulation, we demonstrate that our implementation of canonical correlation analysis performs better than genome-wide analysis on each of the traits individually as well as the principal components of those traits. Our implementation of canonical correlation provides overall less power compared to the established methods of multivariate genome-wide analyses such as multi-trait meta-analysis, and multivariate-PLINK. However, our implementation provides more flexibility by allowing a search for genetic associations in the context of biological processes. ^ We implemented canonical correlation analysis to explore genetically enriched oral glucose tolerance test data for genetic associations. We utilized data from the Starr County Health Studies. This dataset contains an array of traits and biological samples that have been collected over more than three decades. Analyses were performed using 484 Hispanic individuals who underwent oral glucose tolerance test from 2002-2006. We demonstrated a statistically significant canonical correlation between the traits of oral glucose tolerance test and previously documented genetic variants associated with type 2 diabetes. Additionally, we demonstrated a similar relationship with variants representing genes associated with maturity-onset diabetes of young (MODY). Although, in our analyses, no variant reached genome-wide statistical significance using our implementation of canonical correlation analysis, multivariate-PLINK, and multivariate meta-analysis method by Bolormaa et al.[2], our method provides a novel and flexible framework to identify genetically relevant variation while analyzing any set of correlated traits.