A network-based framework for reanalysing public transcriptome data to identify key genes involved in water deficit response in sugarcane (Saccharum spp.)

Background The progress of climate change demands the adoption of crop cultivars more tolerant to water-deficient conditions. Sugarcane ( Saccharum spp.) is a major crop used for the production of sucrose, biomass, and alternative biofuels. Breeding sugarcane cultivars is a complex process, limited by the genetic make-up of the crop. The way around these difficulties depends on the development of biotechnological tools, including genetically modified products. Transcriptome studies provide a reliable source for selecting candidate genes for target modification. However, the choice needs to be effective due to the intrinsic biological challenges of sugarcane. Here, we developed a strategy based on systems biology to demonstrate the reutilization of transcriptomic data in search of central genes and mechanisms involved in response to water deficit, exploring systems biology, co-expression networks studies, and prediction models. Results Sixteen RNA-seq BioProjects of sugarcane grown under contrasting water conditions were retrieved from the SRA-NCBI repository. The 352 libraries were analyzed for differentially expressed genes (DEGs), which were mapped against available sugarcane genomes and assembled transcriptomes. Canonical ontology enrichment and annotation analyses of the combined DEGs allowed the identification of candidate genes, some highly represented and frequently associated with water deficit response, such as inositol 3-alpha-galactosyltransferase. Next, we categorized the 1,415,808 DEGs into 342 functional groups, proposed based on terms commonly related to plant responses to water deficit, and supported by a biological network. All functional groups contain DEGs and their connections; therefore, the member frequencies and number of occurrences among projects allowed us to determine those deviating from normal distribution. We identified 109 functional groups associated with specific pathways that are more active in response to water deficit. The network links enabled the identification of the most central components controlling the mechanisms and pathways. Conclusions The approach proposed here is efficient in providing information on relevant functional groups and the key genes, pathways, and mechanisms associated with those groups. The identification of candidates within the most frequent group effectively demonstrates the relevance of our proposed approach. The strategy offers a robust method for filtering candidates compared to canonical transcriptomic analyses and allows the reuse of data in meta-analyses.