Background: Rapid advances in next-generation sequencing methods have provided new opportunities for transcriptome sequencing (RNA-Seq). The unprecedented sequencing depth provided by RNA-Seq makes it a powerful and cost-efficient method for transcriptome study, and it has been widely used in model organisms and non-model organisms to identify and quantify RNA. For non-model organisms lacking well-defined genomes, de novo assembly is typically required for downstream RNA-Seq analyses, including SNP discovery and identification of genes differentially expressed by phenotypes. Although RNA-Seq has been successfully used to sequence many non-model organisms, the results of de novo assembly from short reads can still be improved by using recent bioinformatic developments.
Results: In this study, we used 212.6 million pair-end reads, which accounted for 16.2 Gb, to assemble the hexaploid wheat transcriptome. Two state-of-the-art assemblers, Trinity and Trans-ABySS, which use the single and multiple k-mer methods, respectively, were used, and the whole de novo assembly process was divided into the following four steps: pre-assembly, merging different samples, removal of redundancy and scaffolding. We documented every detail of these steps and how these steps influenced assembly performance to gain insight into transcriptome assembly from short reads. After optimization, the assembled transcripts were comparable to Sanger-derived ESTs in terms of both continuity and accuracy. We also provided considerable new wheat transcript data to the community.
Conclusions: It is feasible to assemble the hexaploid wheat transcriptome from short reads. Special attention should be paid to dealing with multiple samples to balance the spectrum of expression levels and redundancy. To obtain an accurate overview of RNA profiling, removal of redundancy may be crucial in de novo assembly.