An Integrative Approach to Select Candidate Genes Derived from Sugarcane Transcriptomic Data

Drought and flooding tolerance are desirable breeding characteristics for Colombian sugarcane cultivars due to the high cost of irrigation and drainage systems. Based on a previous identification of genotypes tolerant and susceptible for drought and flooding conditions, a comparative transcriptomics study began as an attempt to identify genes responsive to the stresses of interest. As part of this study, RNA-Seq (Illumina) was used and the transcriptomics data was initially analyzed to identify differentially expressed genes (DEGs). Considering that there is no sequenced sugarcane genome, and thus a consolidated transcriptome, which led to the identification of a high number of DEGs (>4,000 for each study), and the fact that gene expression can be effected by variables non-associated with the condition of interest (drought or flooding stress), the RNA-Seq data was also analyzed to establish relationships among genes where expression values were not prevalent. Among the analyses performed, co-expression networks where created and machine learning algorithm methods, such as support vector machines (SVMs), where implemented. Also, because it was important to identify the genes activated in metabolic pathways modulating process of carbon assimilation, photosynthesis and sucrose synthesis, all the genes were annotated and mapped to the Kyoto Encyclopedia of Genes (KEGG). Finally, in order to consolidate the most important results of each analysis, a selection system, currently involving 19 categories, is being generated and is under construction. The development of this project is a valuable opportunity to apply complex networks and machine learning techniques for gene selection in agriculture, which have not been explored enough within this context. Also, we expect that developed model can be applied in studies in which similar strategies are being used to identify genes of agronomic importance.