IMPROVING GENOMIC ANALYSIS THROUGH JOINT DIMENSION REDUCTION DATA INTEGRATION

Joint dimension reduction (jDR) algorithms are a class of data integration approaches thatsimultaneously reduce the dimensions of multiple datasets, combining them together while preservingthe underlying relationships between them. The use of such methods have immense appeal in clinicalgenomics research, where large, complex datasets are continually being produced and analyzed inorder to identify biological mechanisms underlying disease and inform the development of newtherapies and diagnostic tools. In this work, I show how jDR methods can be leveraged to analyzegenomic and clinical data to gain additional insight into the associations between genomic biomarkersand clinical phenotypes. Further, I demonstrate two extensions of jDR integration to improve thetranslation of these insights from preclinical model systems, such as cell lines or mice, into humanclinical populations. First, through the identification of joint-acting gene variation between model andhuman to improve gene-based biomarkers and predictive models. Second, by identifying joint-actingsample variation, I show how preclinical models could be selected or filtered to better represent aclinical population of interest. Together, the results of the work presented provide a novel perspectiveon the utility of jDR algorithms in genomics research moving forward.