Efficient Computational Genetics Methods for Multiparent Crosses

Multiparent crosses are genetic populations bred in a controlled manner from a finite number of known founders. They represent experimental resources that are of potentially great value for understanding the genetic basis of complex diseases. An important new experimental technology that can be applied to multiparent crosses, namely high-throughput sequencing, generates an immense amount of data and provides unprecedented opportunities to study genetics at a ultra high resolution. However, to take advantage of such massive data, several computational genetics problems have to be resolved. These include RNA-Seq assembly and quantification, QTL mapping, and haplotype effect estimation. In order to tackle these problems, which are highly connected to each other, I propose a series of methods: GeneScissors is a novel method to detect errors caused by multiple alignments in the RNA-Seq; RNA-Skim can rapidly quantify RNA-Seq data while still provide reliable results; HTreeQA is designed as a phylogeny based QTL mapping method for genotypes with heterozygou sites; and Diploffect estimates founder effects with statistically valid interval estimates in multiparent crosses. These methods are extensively studied on both simulated and real data. These studies demonstrate that the proposed methods can make data analysis of multiparent crosses more effective and efficient and produce results are more accurate and trustworthy than a number of existing alternative methods.Doctor of Philosoph