Studying the transcriptome of hexaploid wheat (Triticum aestivum) has been a challenging and time-consuming task until recently. Today, the available reference genome and annotation, the decreasing cost of sequencing, as well as the rapid development of new bioinformatics tools simplify the analysis of the wheat transcriptome dramatically. The next challenge is to maximise the value of the sequencing data for our understanding of the gene regulatory networks underlying traits of agronomical value. Identifying candidate genes among the thousands of transcripts can be very challenging. The work presented here showcases two different approaches to analyse RNA sequencing (RNA-Seq) data and evaluates their strengths and weaknesses for the construction of robust gene networks.

Genie3 is a publicly available algorithm, which was used by Ramirez-Gonzalez et al. (2018) to construct gene regulatory networks for wheat. The open access Genie3 network has the potential to simplify the analysis of transcription networks substantially, however its application in wheat has not yet been experimentally validated. Here we used the network to predict genes downstream of the NAM transcription factors, known to control leaf senescence and grain quality (Uauy et al., 2006). We re-analysed RNA-Seq data from NAM RNA interference and knock-out TILLING mutants (Cantu et al., 2011; Pearce et al., 2014) and showed that a significantly enriched number of Genie3-predicted NAM downstream targets were among the differentially expressed genes in those datasets, suggesting that Genie3 can be used to predict biologically relevant targets.

Parallel to this, we performed a pilot study using RNA-Seq data from extending stem nodes to elucidate the gene regulatory networks and assess potential gene x environment interactions. Although the reduced stature trait conferred by Rht is almost always observed, the expression of the trait can be influenced by environmental factors, such as reduced soil nitrogen availability (Laperche et al., 2007). The RNA-Seq data and automated high-throughput phenotypic data (Phenospex) were collected from Rht near isogenic lines, which were grown in the field under two levels of nitrogen. Based on the Phenospex data, we choose transcriptome data from several developmental time points and will use them to infer gene regulatory networks. We will evaluate the power of combining phenotypic data with causal gene regulatory networks and compare this with the previously described Genie3 network approach. Contrasting advantages and disadvantages of two modelling approaches will enable their further development and more effective use in the future.