A0875
Title: Identifying genes associated with disease outcomes using joint sparse canonical correlation analysis
Authors: Diptavo Dutta - National Cancer Institute (United States) [presenting]
Abstract: Genomic and epigenomic changes can have pivotal effects on cancers, and joint analyses of such multimodal data can identify novel biomarkers for cancer-related outcomes. Joint sparse canonical correlation analysis (jsCCA) is proposed to identify an ensemble of copy number aberrations (CNAs), methylation sites and gene expressions relevant to tumor outcomes. JsCCA detects orthogonal gene modules correlated with sets of methylation sites, which in turn are correlated with sets of CNA. Analysis of data on 515 kidney cancer patients from the TCGA-KIRC found eight gene modules associated with methylation sites and groups of proximally located CNA sites. ASAH1 gene is identified, trans-regulated by CNA and methylation sites, to be associated with tumor stage. Quantifying the overall effect of gene modules revealed that two gene components have significant interaction with smoking and represent distinct biological functions, including inflammatory responses and hypoxia-regulated pathways. The results indicate that methods like jsCCA are warranted for integrative analysis of multimodal data in cancer genomics to identify interpretable, novel, and clinically relevant molecular targets.
