The identification of causal or predictive variants/genes/mechanisms for disease-associated traits is characterized by “complex” networks of molecular phenotypes. Present technology and computer power allow building and processing large collections of these data types. However, the super-rapid data generation is counterweighted by a slow-pace for data integration methods development. Most currently available integrative analytic tools pertain to pairing omics data and focus on between-data source relationships, making strong assumptions about within-data source architectures. A limited number of initiatives exist aiming to find the most optimal ways to analyze multiple, possibly related, omics databases, and fully acknowledge the specific characteristics of each data type. A thorough understanding of the underlying assumptions of integrative methods is needed to draw sound conclusions afterwards. In this chapter, the authors discuss how the field of “integromics” has evolved and give pointers towards essential research developments in this context.
TopIntroduction
DNA and RNA microarray technologies have made it possible to relate genome structure with gene expression patterns and physiological cell states. This paved the way towards a better understanding of tumor development, diseases progression, and drug response (Trevino, Falciani, & Barrera-Saldana, 2007). Since their appearance, these technologies have been used to detect single nucleotide polymorphisms (SNPs) and other structural variations in the genome, such as copy number variations (CNVs) (Feuk, Carson, & Scherer, 2006; Macdonald, Ziman, Yuen, Feuk, & Scherer, 2014; Pang et al., 2010; Pang, Migita, Macdonald, Feuk, & Scherer, 2013), as well as to examine changes involving all aspects of epigenetic interactions (Colyer, Armstrong, & Mills, 2012). Next Generation Sequencing (NGS) can also be used to identify novel mutations. In addition, NGS allows the identification of protein binding to chromatin, RNA quantification, and the investigation of spatial interactions, amongst others. Compared to micro-array experiments, sequencing-based experiments are more widely applicable, since they exhibit a potentially richer information content, but at the expense of higher analytical costs and the need for more sophisticated analytic tools and well-equipped IT-infrastructures to deal with the vast amounts of data they generate (Nekrutenko & Taylor, 2012).
Genome-wide association studies (GWAS) typically assay hundreds of thousands of SNPs in thousands of individuals (Johnson & O'Donnell, 2009). Such studies have reproducibly identified numerous SNP-trait associations, as are catalogued in the National Human Genome Research Institute (NHGRI) Catalog of Published Genome-Wide Association Studies (http://www.genome.gov/gwastudies) (Hindorff et al., 2009). The catalog includes over 1,500 curated publications of over 10,000 SNPs. With the bloom of analytic tools for gene-gene interaction analysis using SNPs (Van Steen, 2012), gene interaction studies are gradually being incorporated in the catalog as well (Welter et al., 2014). However, apart from gene-gene interactions, several other factors exist that makes GWAS less efficient, including compound or multiple phenotypes, genomic imprinting, gene-environment interactions. The latter type of interactions can be taken very broadly, realizing that intermediate phenotypes, such as gene or protein expression, DNA methylation or histone modification, also respond to variations in DNA, cascading into changes for the trait of interest. Clearly, independently carried out omics data analyses are unlikely to be sufficient to obtain a full comprehension of all the underlying principles that govern the functions of biological systems (Joyce & Palsson, 2006).