Genomic Medicine Translation: Evidence Generation for Genomic Risk Assessment, Clinical Implementation, and a New Tool for Variant Interpretation

Abstract

Translating new genomic discoveries into public health benefits requires a concerted, multidisciplinary research effort spanning the translational continuum - from development of candidate health application to evidence-based guidelines, and implementation in practice. This dissertation presents three studies addressing questions related to each phase of this continuum across two genomic medicine applications: population-based genomic risk assessment and variant prioritization for rare disease diagnosis. The first study examined how analytical choices regarding electronic health records (EHR) data affected estimation of genomic risks for coronary heart disease. Results showed that choice of case definition meaningfully influenced relative risk estimations and that specifications of EHR entry time affected estimates of cumulative incidence, suggesting that EHR analytical choice should be explicitly examined in the design and interpretation of genomic risk association studies. The second study characterized barriers and facilitators of implementing a genomic risk assessment in a large health system through key informant interviews. Key barriers identified included evidence uncertainty, complex informational needs, and health system hesitancy to adopt non-guideline recommended interventions, which were partially mitigated by engaging with clinical stakeholders and demonstrating the value of the intervention. The third and final study investigated a novel, interpretable approach for using the deep-learning sequence-to-activity model Enformer to interpret and prioritize non-coding variants underlying rare genetic diseases. Evaluation of the method’s performance in classifying pathogenic and control variants showed that it underperformed conventional scores, highlighting the need for alternative aggregation method to leverage the information predicted by Enformer. Together, these studies generate insights spanning clinical validity, implementation and method development that can inform the future development and implementation of genomic medicine applications.