Videos

Dynamic landmark prediction for genetic mixture models

Presenter
February 21, 2018
Abstract
In kin-cohort studies, clinicians and genetic counselors are interested in providing their patients the most current cumulative risk of a disease arising from a rare deleterious mutation. Estimating the cumulative risk is difficult, however, when the genetic mutation status in patients is unknown and, instead, only estimated probabilities of a patient having the mutation are available. We propose to estimate the cumulative risk for this scenario using a novel nonparametric estimator that incorporates covariate information and dynamic landmark prediction. Our contributions are three-fold. Our estimator is shown to better inform patients of their disease risk as it yields improved prediction accuracy over existing estimators that ignore covariate information. Our estimator is built within a dynamic landmark prediction framework whereby we can obtain personalized dynamic predictions over time. Lastly, compared to current standards, a simple transformation of our estimator provides more efficient estimates of marginal distribution functions in settings where patient-specific predictions are not necessarily the main goal. Using simulation studies, we show our estimator is unbiased and has significant gains in predictive accuracy compared to approaches that ignore covariate information and landmarking. Applying our method to a mortality kin-cohort study of Huntington's disease, we show that incorporating familial genetic information and updated survival information yields more accurate estimates of survival for individuals at risk for the disease compared to methods that do neither.