Annotation-informed integration of 'omic data in next-generation sequencing

Douglas D Baumann, Purdue University

Abstract

Next-generation sequencing (NGS) technologies have opened the door to a wealth of knowledge and information about biological systems, particularly in genomics and epigenomics. These tools, although useful, carry with them additional technological and statistical challenges that need to be understood and addressed. Four separate, but related, NGS issues are addressed in this work. Specifically, we address amplification bias, an improvement in statistical power for differential methylation tests, boundary definitions and analyses for histone modifications, and the integration of genomic and epigenomic data in joint analyses. The majority of NGS technologies effectively sample small amounts of DNA or RNA that are amplified (i.e., copied) prior to sequencing. The amplification process is not perfect, and thus sequenced read counts can be extremely biased. Unfortunately, current amplification bias controlling procedures introduce a dependence of gene expression on gene length. This issue is addressed via a novel procedure that accounts for amplification bias, and is effective in estimating true gene expression independent of gene length. NGS technologies generate whole-genome epigenomic profiles at single base-pair resolution. Several novel statistical approaches are presented that incorporate genomic annotation information into the analysis of differential methylation and histone modification profiles for the purpose of increasing the statistical power. A novel application of hierarchical Bayesian modeling, that incorporates DNA methylation information into a differential gene expression testing framework, is also presented. The advantages of these approaches are demonstrated through extensive simulations and real data applications.

Degree

Ph.D.

Advisors

Doerge, Purdue University.

Subject Area

Statistics|Bioinformatics

COinS