Increasing Inclusion of Participants, Leveraging Gut Microbiomes, and Optimizing Machine Learning Models to Advance Warfarin Precision Medicine

Abstract

Precision medicine, one of the promises of the future of clinical medicine, will not be realized without challenges. Here, focusing on an individual pharmaceutical therapeutic, warfarin, we demonstrate idealistic benefits, realistic caveats, and paths to improving the benefits and equity of precision medicine through rigorous research. Anticoagulation therapy with the popular vitamin K antagonist, warfarin, can be challenging due to large interpatient dose variability and a narrow therapeutic window. Warfarin genotype-guided dose prediction models currently can account for approximately 50% of the total variability seen across patients, however, this number can also be as low as 20% in diverse populations. This lack of equity and overall dismal performance of genotype-guided dosing has reasonably slowed the clinical implementation of warfarin dose predictions models and is a driving force of the research projects presented here. The paths that we suggest for improving future benefits of precision medicine research: 1) consciously include everyone, 2) survey the environment, regardless of temporal stability, and 3) validate the generalizations we make. In this thesis we briefly introduce the fields of pharmacogenetics, pharmacomicrobiomics, and statistical modeling to provide evidence for each of the previous suggestions demonstrated by the attached appendices. Finally, we make predictions regarding the future of warfarin precision medicine.