Similarity network fusion for the integration of multi-omics and microbiomes in respiratory disease

Abstract

Advances in platform technologies facilitate the design of largescale “multi-omic” studies that encompass genomic, transcriptomic, proteomic, epigenomic, metabolomic and microbiomic components, each representing different views of a single biological specimen [1]. While useful, this is analogous to the ‘Flatland’ jeu d'esprit where the same reality (i.e., a sphere of constant diameter) is subject to different interpretations (i.e., circles of varying diameter) depending on one’s point of view (from various 2-D cross sections). Although each -omics approach has value, they can be even more useful if holistically modelled through appropriate integration. While ‘mono-omic’ analysis has been extremely beneficial, from a systems medicine perspective, this may fail to capture the emergent properties of an individual system and hence may yield limited understanding of non-linear and dynamic features, all of which are increasingly evident in the pathogenesis of respiratory disease [1]. There is clearly a growing need for a more holistic ‘all in’ integration methodology that leverages each distinct -omic dataset derived from multi-omic studies (Figure 1). Although several integrative methodologies are available (e.g. mixOmics, Anvi'o and integrOmics), Similarity Network Fusion (SNF) has emerged as an appropriate, applicable, and robust method in respiratory disease [2-4].