Computational approaches for analyzing variation in the transcriptome and methylome of Zea mays

Abstract

While Mendelian genetic approaches to crop improvement have been successful in the past, effective modern breeding programs are becoming increasingly dependent on accurate information about gene functionality and regulatory mechanisms. Recent advances in sequencing technologies have produced the complete genomes of many organisms, but the annotation of predicted genes still lags behind. Since domesticated varieties tend to be phenotypically divergent from their ancestral species, the examination of domestication effects on their transcriptomes can provide beneficial insights into the function of genes targeted during domestication.This dissertation focuses on computational approaches for comparative analysis of gene expression, which is a valuable resource for gene annotation. We begin with the analysis of two co-expression networks built on expression data from maize and its wild ancestor, teosinte. We reveal biologically significant differences between the two networks and propose a novel method to identify genes with altered expression covariation between the two species. We show that our approach is more sensitive than existing methods and illustrate its complementarity to differential expression or genome sequence analysis. The approach is also applied to study differences between networks derived from RNA-seq and microarray gene expression data, where we identified and resolved issues with comparing and combining co-expression networks derived from the two data types.In the second part of the dissertation, we describe a pipeline for the identification of differentially methylated regions in maize and teosinte. Application of this approach to a diverse set of maize lines suggests the presence of purely epigenetic alleles and confirms the prevalence of the negative relationship between DNA methylation and the expression levels of nearby genes.