Gene selection And hub gene identification Using multiomics data

Abstract

Gene selection or biomarker discovery is the process of identifying a subset of newlinerelevant and informative genes from the original set of genes with diagnostic and newlineprognostic capability. Their discriminative ability allows classifying samples into newlinedisease categories (diagnostic), while their predictive power enables assessing the newlinecause of disease and discovery of new therapy (prognostic). Even though the DNA newlinemicroarray technology has given the researchers, a remarkable opportunity to newlineanalyze the expression pattern or genetic signature of thousands of genes newlineconcurrently in a solitary platform, still it is limited with large dimension, high newlinenoise, batch effect, and low reproducibility. Technological advances in high newlinethroughput sequencing technology generate a plethora of data from multiple levels newlineof biological systems such as genome, epigenome, transcriptome, proteome, and newlinemetabolome, which is collectively called as multiomics data. newlineConventional approaches of gene selection mostly rely on analyzing the newlinedifferent types of omics data at a single level that focuses on identifying the newlinevariations at a single level, at the same time neglecting the causal relationship newlinebetween multiple levels of biological entities. It results in a poor understanding of newlinedisease pathogenesis and partial exploration of malignant transformation. In newlinecontrast, integrative approaches combine data from multiple oms which give a newlinedeeper insight into the interplay of molecules across omics levels and thereby newlinebridge the gap between genotype to phenotype. newlineIn this work, we aim at outstretching the existing machine learning-based gene newlineselection approaches by acclimatizing network-based gene selection using multiple newlinelevels of omics data. The objective of this thesis is to perform gene selection and newlineidentify few signature hub genes in the genetic network, which are statistically newlineviii newlineAbstract newlinecompetent and biologically enriched. The purpose of this thesis is to address the newlinechallenges of the high dimensionality problem, issues in analyzing the single level