The main research question: What is the current and future predictive accuracy of polygenic scores? The first aim contributes to better polygenic scores and studies their relationship with environmental risk factors. The second aim estimates the accuracy of scores with larger genome-wide association studies, followed by simulation of the disease risk distributions that future polygenic scores could eventually predict. A harmonized resource of risk distributions for a wide selection of medical conditions will be returned to help accelerate research on polygenic scores.
The third aim applies the distributions to study the usefulness of integrating polygenic scores in disease prediction models, e.g., those offered by consumer genetics services. Performance metrics like accuracy, cost-effectiveness, personal utility, and optimal uptake will be estimated, also conditional on actionability. The performance and limitations of using polygenic scores for health-contingent decision making (e.g., insurance or retirement planning), or personal advice (e.g., preventive lifestyle changes), will be investigated. The results will be evaluated to inform the design of public policy-including public health policy, consumer protection law, and insurance regulation that seeks to balance efficiency, fairness, and privacy concerns from the increasing accuracy and availability of polygenic scores.
Third aim details and data. For breast cancer, we used UKB micro data, previously estimated PRS coefficients, and heritability estimates to estimate our model. The model then gives a joint distribution of age, breast cancer risk based on age alone, and based on combinations of covariates and current and future PRS. We use this joint distribution to estimate the gains from alternative screening guidelines using PRS.