Date of Award
Doctor of Philosophy
Advances in technologies currently produce more and more cost-effective, high-throughput, and large-scale biological data. As a result, there is an urgent need for developing efficient computational methods for analyzing these massive data. In this dissertation, we introduce methods to address several important issues in gene expression and genomic sequence analysis, two of the most important areas in bioinformatics.Firstly, we introduce a novel approach to predicting patterns of gene response to multiple treatments in case of small sample size. Researchers are increasingly interested in experiments with many treatments such as chemicals compounds or drug doses. However, due to cost, many experiments do not have large enough samples, making it difficult for conventional methods to predict patterns of gene response. Here we introduce an approach which exploited dependencies of pairwise comparisons outcomes and resampling techniques to predict true patterns of gene response in case of insufficient samples. This approach deduced more and better functionally enriched gene clusters than conventional methods. Our approach is therefore useful for multiple-treatment studies which have small sample size or contain highly variantly expressed genes.Secondly, we introduce a novel method for aligning short reads, which are DNA fragments extracted across genomes of individuals, to reference genomes. Results from short read alignment can be used for many studies such as measuring gene expression or detecting genetic variants. Here we introduce a method which employed an iterated randomized algorithm based on FM-index, an efficient data structure for full-text indexing, to align reads to the reference. This method improved alignment performance across a wide range of read lengths and error rates compared to several popular methods, making it a good choice for community to perform short read alignment.Finally, we introduce a novel approach to detecting genetic variants such as SNPs (single nucleotide polymorphisms) or INDELs (insertions/deletions). This study has great significance in a wide range of areas, from bioinformatics and genetic research to medical field. For example, one can predict how genomic changes are related to phenotype in their organism of interest, or associate genetic changes to disease risk or medical treatment efficacy. Here we introduce a method which leveraged known genetic variants existing in well-established databases to improve accuracy of detecting variants. This method had higher accuracy than several state-of-the-art methods in many cases, especially for detecting INDELs. Our method therefore has potential to be useful in research and clinical applications which rely on identifying genetic variants accurately.
dissertation or thesis originally submitted to the local University of Memphis Electronic Theses & dissertation (ETD) Repository.
Vo, Nam Sy, "Computational Methods for Gene Expression and Genomic Sequence Analysis" (2016). Electronic Theses and Dissertations. 1457.