A Computational Framework for Scalable Epistasis Analysis on High-Dimensional Genomic Data.

Shi, Xinghua

Abstract

Data tsunami in genomic medicine calls for robust and scalable methods for genome scale analysis of a large number of genetic variants regarding to their complicated impact on various human traits. A systematic analytical framework in this study presents a new opportunity to rigorously assess the low and high order effect of numerous genetic variants on a wide range of traits.
The specific aims of our proposed project are as follows:
Aim 1. Develop covariate-aware models for pairwise epistatic analysis on a categorical trait;
Aim 2. Develop scalable methods for multilocus and multi-trait epistatic analysis;
and Aim 3. Extensively evaluate the methods using genomic datasets and deploy a computational frame- work with a Web-portal service. This study has several innovations: 1) we will provide a ?one-stop-shop? for comprehensive epistasis analysis of large-scale genomic datasets; 2) we will develop robust and scalable methods for dissecting main and epistatic effect of genetic variants on one or more traits of different types; 3) we will apply these methods to a variety of genomic datasets at different scales. The framework from our study will be flexible to include prospective large- scale genomic and epigenomic datasets in the future.

Public Health Relevance

(Public Health Relevance Statement) We propose a systematic and comprehensive analytical study that aims to identify high order relationship among genetic variants, i.e. multilocus epistasis, regarding their joint effect on various traits of interest. In this study, we will provide a computational framework that allows a one-stop shop for complex analytics of mining high dimensional genomic datasets. The proposed research is relevant to public health and the mission of NIH because the accomplishment of our proposed work is expected to facilitate the identification of genetic variants underlying various traits including human diseases, and help us to understand, prevent, diagnose, and treat these diseases.