Polygenic scores (PS) represent an exciting translational application of genomic information that can mitigate some short comings of traditional genetic approaches, and which have great potential towards understanding and leveraging genetic predisposition to health traits. However, key barriers to clinical application exist, prominently including the fact that current PS were developed nearly exclusively using European individuals. These PS are much less predictive when applied to non-European populations (e.g. African Americans). Second is that there has been only very limited application of PS to pharmacogenomics, despite that medication use is arguably the most common and impactful type of medical intervention for patients with common chronic diseases. The goal of this proposed project is to develop novel methods to create and validate PS for African Americans that predict medication response in common complex conditions. To accomplish the goal of valid and ancestry robust PS for African Americans, we will develop innovative analytic approaches that incorporate localized ancestry to construct PS in a manner that allows for SNP selection and the allelic effect to be dependent on the local ancestry context, thus creating tools that are robust to varying genetic admixture and population structure. We will then compare the performance of these novel tools to two current methods (including state-of-the-art Bayesian approaches) within cohorts of African American patients across two different common complex diseases (heart failure and diabetes).