Combining independent de novo assemblies optimizes the coding transcriptome for nonconventional model eukaryotic organisms

Abstract

BackgroundNext-generation sequencing (NGS) technologies are arguably the most revolutionary technical development to join the list of tools available to molecular biologists since PCR. For researchers working with nonconventional model organisms one major problem with the currently dominant NGS platform (Illumina) stems from the obligatory fragmentation of nucleic acid material that occurs prior to sequencing during library preparation. This step creates a significant bioinformatic challenge for accurate de novo assembly of novel transcriptome data. This challenge becomes apparent when a variety of modern assembly tools (of which there is no shortage) are applied to the same raw NGS dataset. With the same assembly parameters these tools can generate markedly different assembly outputs.ResultsIn this study we present an approach that generates an optimized consensus de novo assembly of eukaryotic coding transcriptomes. This approach does not represent a new assembler, rather it combines the outputs of a variety of established assembly packages, and removes redundancy via a series of clustering steps. We test and validate our approach using Illumina datasets from six phylogenetically diverse eukaryotes (three metazoans, two plants and a yeast) and two simulated datasets derived from metazoan reference genome annotations. All of these datasets were assembled using three currently popular assembly packages (CLC, Trinity and IDBA-tran). In addition, we experimentally demonstrate that transcripts unique to one particular assembly package are likely to be bioinformatic artefacts. For all eight datasets our pipeline generates more concise transcriptomes that in fact possess more unique annotatable protein domains than any of the three individual assemblers we employed. Another measure of assembly completeness (using the purpose built BUSCO databases) also confirmed that our approach yields more information.ConclusionsOur approach yields coding transcriptome assemblies that are more likely to be closer to biological reality than any of the three individual assembly packages we investigated. This approach (freely available as a simple perl script) will be of use to researchers working with species for which there is little or no reference data against which the assembly of a transcriptome can be performed.Electronic supplementary materialThe online version of this article (doi:10.1186/s12859-016-1406-x) contains supplementary material, which is available to authorized users.

The Holo-Transcriptome of a Calcified Early Branching Metazoan

“…Read qualities were assessed using FastQC (Andrews, 2014), and Trimmomatic (Bolger et al, 2014) was used to remove low quality reads and adapter sequences. Reads were assembled de novo using our recently developed assembly pipeline (Cerveau and Jackson, 2016). Briefly, we employed three assembly packages with unique assembly strategies: Trinity V2.0.3 (Haas et al, 2013), CLC Genomics Workbench (www.qiagenbioinformatics.com) and IDBA-tran V1.1.1 (Peng et al, 2013).…”

“…Quality filtering of raw reads resulted in 34,592,975 paired end reads that were assembled as described by Cerveau and Jackson (2016), see Methods section. The assembly generated 91,443 contigs with an average length of 933 bp and an N50 value of 1,307 ( Table 1).…”

Trial by phylogenetics - Evaluating the Multi-Species Coalescent for phylogenetic inference on taxa with high levels of paralogy (Gonyaulacales, Dinophyceae)

“…Further, Trinity performed well for identifying isoforms of genes and excelled at assembling highly expressed genes (Honaas et al, 2016). Conversely, Cerveau and Jackson (2016) found that Trinity, CLC Bio and IDBA-Tran assemblies all contain errors introduced by the assembly algorithms. Using a combination of all three assemblers yielded a final assembly closer to biological reality than any individual assembler, when no reference genome is available (Cerveau and Jackson, 2016).…”

“…Using a combination of all three assemblers yielded a final assembly closer to biological reality than any individual assembler, when no reference genome is available (Cerveau and Jackson, 2016). As our present study used Trinity exclusively, it may be subject to the type of errors found by Cerveau and Jackson (2016) which could affect downstream analysis.…”

Gonad transcriptome of golden musselLimnoperna fortuneireveals potential sex differentiation genes

“…The alignments generated by the HISAT2 were then used to assemble the transcriptomes using StringTie or Trinity. To reduce redundancy while keeping the relevant biological information we used the script concatenator.pl (34) with the three separate assemblies as input. The script performed intra-assembly clustering with CD-HIT-EST (35), the unique transcripts from the assemblies were concatenated, and ORFs were predicted using TransDecoder ver.3.0.1 (http://transdecoder.github.io).…”