Date of Award
Doctor of Philosophy (PhD)
Computational Biology and Bioinformatics
Genetic prediction of complex traits, also known as polygenic risk score (PRS), is constructed by combining the estimated effect sizes of genetic markers across the genome for an individual. PRS has shown great promise in biomedical and clinical research for disease prevention, monitoring and treatment. However, the development of accurate prediction models is challenging due to the wide diversity of genetic architecture, limited access to individual level data, and the demand for computational resources. The broader application of PRS to the general population is further hindered by the poor transferability of PRS developed in Europeans to non-European populations. In this thesis, we develop two statistical methods to help address these limitations. Chapter 1 includes a review of PRS from a statistical perspective. In Chapter 2, we present a summary statistics-based nonparametric method SDPR that is adaptive to different genetic architectures, statistically robust, and computationally efficient. The material is drawn from the manuscript “A fast and robust Bayesian nonparametric method for prediction of complex traits using summary statistics” with minor modification. In Chapter 3, we develop a statistical method called SDPRX that can effectively integrate genome wide association study summary statistics from different populations to improve the prediction accuracy in non-European populations. The material is drawn from the manuscript “SDPRX: A statistical method for cross-population prediction of complex traits” in preparation.
Zhou, Geyu, "Statistical Methods for Genetic Prediction of Complex Traits in Single and Multiple Populations" (2022). Yale Graduate School of Arts and Sciences Dissertations. 690.