AbstractSystems-level analyses, such as differential gene expression analysis, co-expression analysis, and metabolic pathway reconstruction, depend on the accuracy of the transcriptome. Multiple tools exist to perform transcriptome assembly from RNAseq data. However, assembling high quality transcriptomes is still not a trivial problem. This is especially the case for non-model organisms where adequate reference genomes are often not available. Different methods produce different transcriptome models and there is no easy way to determine which are more accurate. Furthermore, having alternative splicing events could exacerbate such difficult assembly problems. While benchmarking transcriptome assemblies is critical, this is also not trivial due to the general lack of true reference transcriptomes. In this study, we provide a pipeline to generate a set of the benchmark transcriptome and corresponding RNAseq data. Using the simulated benchmarking datasets, we compared the performance of various transcriptome assembly approaches including genome-guided, de novo, and ensemble methods. The results showed that the assembly performance deteriorates significantly when the reference is not available from the same genome (for genome-guided methods) or when alternative transcripts (isoforms) exist. We demonstrated the value of consensus between de novo assemblers in transcriptome assembly. Leveraging the overlapping predictions between the four de novo assemblers, we further present ConSemble, a consensus-based de novo ensemble transcriptome assembly pipeline. Without using a reference genome, ConSemble achieved an accuracy up to twice as high as any de novo assemblers we compared. It matched or exceeded the best performing genome-guided assemblers even when the transcriptomes included isoforms. The RNAseq simulation pipeline, the benchmark transcriptome datasets, and the ConSemble pipeline are all freely available from: http://bioinfolab.unl.edu/emlab/consemble/.Author summaryObtaining the accurate representation of the gene expression is critical in many analyses, such as differential gene expression analysis, co-expression analysis, and metabolic pathway reconstruction. The state of the art high-throughput RNA-sequencing (RNAseq) technologies can be used to sequence the set of all transcripts in a cell, the transcriptome. Although many computational tools are available for transcriptome assembly from RNAseq data, assembling high-quality transcriptomes is difficult especially for non-model organisms. Different methods often produce different transcriptome models and there is no easy way to determine which are more accurate. In this study, we present an approach to evaluate transcriptome assembly performance using simulated benchmarking read sets. The results showed that the assembly performance of genome-guided assembly methods deteriorates significantly when the adequate reference genome is not available. The assembly performance of all methods is affected when alternative transcripts (isoforms) exist. We further demonstrated the value of consensus among assemblers in improving transcriptome assembly. Leveraging the overlapping predictions between the four de novo assemblers, we present ConSemble. Without using a reference genome, ConSemble achieved a much higher accuracy than any de novo assemblers we compared. It matched or exceeded the best performing genome-guided assemblers even when the transcriptomes included isoforms.
De Novo Transcriptome Assembly, Functional Annotation and SSR Marker Discovery of Qinling Takin (Budorcas taxicolor bedfordi)