Parallel integration of heterogeneous genome-wide data sources

Abstract

Heterogeneous genome-wide data sources capturetl information on various aspects of complex biological systems.For instance, transcriptome, interactome and phenome-levelinformation may be derived from mRNA expression data,protein-protein interaction networks, and biomedical literaturecorpora. Each source provides a distinct "view" of the sameomain, but potentially encodes different biologically-relevantpatterns. Effective integration of such views can provide aricher, more informative model of an organism's functionalmodules than that produced on a single view alone. Existingmachine learning strategies for information fusion largely focuson the production of a consensus model that reflects patternsshared between views. However, the information provided by different views may not always be easily reconciled, due tothe incomplete nature of the data, or the fact that someinpatterns will be present in one view but not in another.To address this problem, we present the Parallel IntegrationClustering Algorithm (PICA), a novel cluster analysis approach which supports the simultaneous integration of informationfrom two or more sources. The resulting model preservespatterns that are unique to individual views, as well as those common to all views. We demonstrate the effectiveness of PICA in identifying significant patterns corresponding to functionalgroupings, when applied to three genome-wide datasets.