A cross-platform comparison of transcriptome de novo assemblers using the non-model plant Pelargonium x hortorum (Geraniaceae)

Transcriptome sequencing (RNA-seq) on non-model organisms is becoming increasing popular and has been applied in many areas of biological research. The wide use of RNA-seq and various sequencing platforms and assemblers calls for a thorough comparison of the most popular assemblers for the different platforms in terms of data quality, cost and time. A common problem of de novo transcriptome sequencing on non-model organisms is the lack of a closely related reference genome, making the reconstruction and evaluation of the transcriptome assembly challenging. We have conducted a cross-platform comparison of Pelargonium x hortorum transcriptome assembly using the Illumina and Roche/454 sequencing platforms with five different assemblers: Newbler, MIRA, SOAPdenovo, Trinity and SOAPdenovo-Trans (SOAPtrans). Approximately 500 thousand 454 sequencing reads and 46 million Illumina reads were generated. Reads were postprocessed and then assembled with platform-specific software programs. Using basic metrics such as total assembly size and mean contig and maximal contig lengths, we found that Trinity performed significantly better than other assemblers. Two previously established parameters, contiguity and completeness, were also calculated by comparing the assembly results from the five assemblers to a custom data set of Arabidopsis nuclear proteins that are targeted to plastid, mitochondrion, or cytosol. Trinity and SOAPtrans had significantly higher contiguity and completeness than the assemblies from the other three programs. Common assembly issues (redundancy, fragmentation i.e.) are also discussed. The combination of Illumina sequencing and Trinity or SOAPtrans assembly gave the best results in terms of the basic assembly metrics and generated the most complete transcriptome among all the assemblers analyzed. Thus, we recommend the Illumina/Trinity or SOAPtrans approach as the optimal strategy for transcriptome sequencing and de novo assembly for non-model organisms. This comparison will help resolve the controversy and confusion regarding the selection of the best platform and assembler for transcriptome de novo assembly in situations where reference-based assembly is not possible.