Precision Phenomics to Personalize Drug Therapy The promise of genomic medicine is the personalization of therapeutics based on one's genetic makeup. Current methods to identify genomic variation underlying adverse drug reactions (ADRs) and to predict drug effects progress slowly and lack disease-neutral approaches. In addition, the cost of new drug discovery is increasing rapidly, largely due to ADRs and efficacy failures. Thus, repurposing old drugs for new indications offer advantages of both safety and reduced development costs. We will develop and apply phenome-wide, medication-wide approaches to dense, longitudinal Electronic Health Record (EHR) data linked to DNA to discover genetic variants associated with ADRs, predict new indications for drugs, and identify new phenotype associations for genetic variants known to impact drug response.
In Specific Aim 1, we will genotype 30,000 individuals on a genome-wide array enriched with variants underlying cardiac electrical activity, drug metabolism, HLA variants, and drug targets. Combining these data with the extant genotypes in our institutional biobank BioVU will result in population of 66,000 individuals with dense genome-wide genotype data. We will perform phenome-wide associations studies (PheWAS, a methodology we have developed) for genes and variants identified in Projects 1 and 2 and known pharmacovariants.
In Specific Aim 2, we will use phenome-wide approaches to repurpose existing medications and predict side effects. Building on methods that successfully replicated known apremilast (PDE4 inhibitor used for autoimmune disease) indications and suggested new, biologically plausible repurposing in other diseases, we will perform PheWAS on drug targets for nearly all currently used medications as a tool to identify new disease indications and side effects. Existing indications will serve as anchors to orient results toward new efficacies and possible side effects. Then, we will prioritize new indications for further analysis using network analysis and systematic evidence reviews.
In Specific Aim 3, we will use natural language processing and coded EHR data to identify ADRs from EHR data. Specific ADRs assessed will include diseases, laboratory abnormalities, cutaneous hypersensitivity reactions, and electrocardiographic traits. Our methods will extract both provider-identified ADRs as well as find known clinical events documented but not explicitly recorded as an ADR. Then, we will discover genetic variants predicting the ADRs. In both Specific Aims 2 and 3, we will replicate prioritized novel associations in external EHR-linked biobanks and using candidate gene sequencing or specific HLA 4-digit typing of validated phenotypes. The results of this study will be to dramatically increase the catalog of genetic predictors of drug response and to create a library of potential repurposing for nearly all medications.
Showing the most recent 10 out of 114 publications
