Integrated Analysis of Multiple Data Sets With Biomedical Applications

Integrated Analysis of Multiple Data Sets With Biomedical Applications Public Deposited

Last Modified

Creator

Li, Gen
- Affiliation: College of Arts and Sciences, Department of Statistics and Operations Research

Abstract

It is increasingly common to have measurements from multiple platforms on the same set of samples in modern biomedical sciences. In this dissertation, we develop novel methodologies for integrated analysis of multiple data sets. In particular, we devise a supervised principal component analysis framework that achieves dimension reduction of the primary data with guidance from an auxiliary data set. It extracts accurate and interpretable low-rank structures that are potentially driven by the auxiliary information. We further extend the method to accommodate special features of data such as functionality and high dimensionality through regularization. Numerical examples demonstrate that the proposed methodologies have clear advantages over existing methods. In addition, we develop a Bayesian hierarchical model for multi-tissue eQTL analysis. It exploits shared information in multiple tissues to increase the power of eQTL discovery and improve tissue specicity assessment. The method has been adopted by the Genotype-Tissue Expression (GTEx) consortium and successfully applied to the nine-tissue pilot data.

Date of publication