A multiway approach to data integration in systems biology based on Tucker3 and N-PLS

This paper discusses the potential of multi-way projection methods for analysing multifactorial data structures to identify underlying components of variability that interconnect different blocks of omics variables. We explore their suitability for explorative and variable selection analysis of systems biology data where different types of biological parameters are studied together. These methodologies were applied to the integrative analysis of a functional genomics dataset where transcriptomics, metabolomics and physiological data are available. Our results show that multiway methods are suited to accommodate multifactorial omics experiments and to analyse relationships between different biochemical layers. Additionally, strategies are presented for variable selection in the context of omics data and for interpreting results at the level of cellular pathways.