| dc.description.abstract | Beginning in the 2000s, electronic health record (EHR) systems have collected large
amounts of clinical data. To leverage EHR data for drug studies, researchers have developed powerful computational methods. However, these methods have often been difficult to transport between databases due to technological barriers. Here, I describe two portable methods using EHR data to detect drug-drug interaction (DDI) signals and to validate drug repurposing candidates. First, I developed a method called drug-drug interaction wide association study (DDIWAS). DDIWAS uses information from the common EHR allergy list to validate known DDIs and identify potential novel DDIs. DDIWAS was used to detect DDIs for two common drugs, simvastatin and amlodipine. Second, I developed a method integrating human phenotype transcriptomic signatures, drug perturbation data, and EHR data to validate drug repurposing candidates. This method is transportable, because it uses publicly available data from large human genetic studies and clinical data from the National Institutes of Health All of Us Research Program. Further, unlike existing methods, this approach allows researchers to compare the treatment effect of drug repurposing candidates to known approved drugs in real- world patients. This method was applied to screen for and clinically validate drug repurposing candidates for hyperlipidemia and hypertension. Software packages implementing these methods are both publicly available on GitHub. In summary, work in this thesis provides novel portable methods leveraging EHR data to identify DDI signals and find new uses for existing drugs. | |
