Systematic and Integrative Analysis of Gene Expression to Identify Feature Genes Underlying Human Diseases

Abstract

Over the past two decades, the advances in genomics technology have opened the door for rapid biological data acquisition and have revolutionized many aspects of biomedical research. Given the complex and noisy nature of the large-scale biological data, there is a high demand for developing variable selection approaches to identifying disease biomarkers in the field of translational bioinfor- matics. These biomarkers offer early detection of pathogenesis, inform prognosis, provide guidance for the treatment, and monitor disease progresses. In this chapter, we focused on developing a variety of methods that systematically analyzed whole-genome gene expression data for identifying feature genes associated with patient clinical parameters. In the first method, we constructed a gene co-expression network and then selected genes that are informative for classifying different cancer subtypes based on gene connectivity within the co-expression network. In the second method, we incorporated prior biological pathway information to recon- struct a gene network and then identified hub genes that are associated with cancer prognosis. Finally, we identified protein subnetworks instead of individual genes as biomarkers for classifying different types of brain injuries. Our study has set up a framework that can be easily generalized to integrate different types of genomics and proteomics information for better identifying feature genes to improve accuracy of disease diagnosis and treatment.