Gene Regulatory Network Discovery Through Data Mining: Proof of Concept Through Known Stress Pathway Recapitulation.

The flood of data pouring from increasingly affordable omics platforms has presented a major challenge to existing analytical approaches. As the resulting pool of private and public data deepens, it also presents new opportunities to ask questions not originally envisioned during the generation of individual datasets. Just as earlier molecular approaches evolved from the characterization of individual molecules to entire “omes,” we expect that it will become increasingly valuable to combine disparate datasets, created by different researchers for different purposes, in order to form custom in silico experiments. Taking advantage of existing datasets will be especially useful for the identification of genes involved in complex, environmentally and developmentally specific traits, where the collection of high quality data remains difficult. In order to further advance our current ability to identify candidate genes associated with traits of interest, we investigated to what extent separate experiments, conducted on a shared platform, could be harmonized to create transitive gene transcription meta-networks. To do this, we relied upon individual datasets that each included a time component, thus allowing for the characterization of dynamic associations among molecules in each individual network prior to assembling the meta-network. Such an approach is especially relevant to the challenge of developing genetic resistance to biotic and abiotic stresses, which develop in an environment and tissue specific manner. We applied this approach successfully to recapitulate known developmentally-specific stress response pathways from existing data.