Article Preview
TopIntroduction
In this paper, a powerful gene analytics web application is introduced, which overall combines computational methodologies and data visualization techniques, in order to deliver comprehensible illustrations of cellular complexity, for voluminous, high dimensional, molecular datasets, that address genome in its functional entirety. Moreover, by linking the individual genes, with the critical cellular processes that orchestrate physiological response, through the exploitation of appropriate controlled vocabularies, which ideally possess a tree structure and aid application of deductive logic and knowledge discovery, it achieves a systemic interpretation of the mechanism interrogated, while at the same time it manages to prioritize functionally genes with a pronounced regulatory involvement in the important cellular processes underlying the phenotype studied.
The completion of Human Genome Project in April 2003 (Venter et al., 2001) marks the dawn of the genomic era, disclosing the astonishing complexity of the genomic regulation and functionality that underlies phenotypic diversity. Ever since, dramatic scientific progress has been made, delivering advanced experimental and analytical methodologies that leverage the noted complexity and aid the elucidation of intricate biological problems. In particular, advancements in DNA microarray (Gresham, Dunham, & Botstein, 2008), or more recently next generation sequencing technologies (Goodwin, McPherson, & McCombie, 2016) have generated an avalanche of high-throughput experimental genomic data, holding the promise for the decipherment of complex regulatory mechanisms, underlying the genesis and progression of complex diseases or more generally pathologies, through the creation of stable links between the comprehension of aberrant homeostatic response and the phenotypic cues pertinent with disease manifestation. Moreover, the development of several biomedical ontologies and databases, structuring and categorizing the knowledge amassed concerning the study of biomolecules in different temporal, spatial and organizational scales, has contributed to the rapid expansion of the digitized informational universe, through the avid accumulation of semantic terms, describing biological complexity of inter- and intra-cellular mechanisms (Ashburner et al., 2000), encompassing metabolic and signaling pathways (Croft et al., 2014; Kanehisa & Goto, 2000) or phenotypes (Eppig et al., 2014; Köhler et al., 2014) among others. This highly heterogeneous and therefore idiosyncratic ecosystem, poses significant challenges, when it comes to its efficient integration and meaningful interpretation into concrete, mechanistic modeling approaches. This enables a deeper understanding of the mechanism under investigation on one hand, while at the same time provides a rational framework for the testing of novel hypotheses, as proposed from a computational, data-driven research perspective.
High-throughput and large-scale genomic analysis leads to extended lists of genes, with differentiated status among the various phenotypic categories observed, or more broadly the examined stratifications. These profiles set the genomic fingerprint of the examined phenotypes. Statistical enrichment analysis (Subramanian et al., 2005) is an efficient strategy to systematically associate the extracted gene list with various functionalities, as these are encapsulated in terms of biomedical vocabularies, revealing the respective mechanistic impression of a biological phenotype at the given level of molecular dissection that the vocabulary describes. Various enrichment analysis approaches exploit biomedical ontologies (vocabularies with hierarchical tree structure) and gene repositories integrating mappings between genes and functional or structural descriptions, through the application of well-known statistical methods (Fisher’s exact test, Hypergeometric distribution, Chi-squared test) with the scope to rigorously reveal the over-represented entities, correlated with the input list (Huang, Sherman, & Lempicki, 2009).