Cutoffs and k-mers: implications from a transcriptome study in allopolyploid plants

Abstract

BackgroundTranscriptome analysis is increasingly being used to study the evolutionary origins and ecology of non-model plants. One issue for both transcriptome assembly and differential gene expression analyses is the common occurrence in plants of hybridisation and whole genome duplication (WGD) and hybridization resulting in allopolyploidy. The divergence of duplicated genes following WGD creates near identical homeologues that can be problematic for de novo assembly and also reference based assembly protocols that use short reads (35 - 100 bp).ResultsHere we report a successful strategy for the assembly of two transcriptomes made using 75 bp Illumina reads from Pachycladon fastigiatum and Pachycladon cheesemanii. Both are allopolyploid plant species (2n = 20) that originated in the New Zealand Alps about 0.8 million years ago. In a systematic analysis of 19 different coverage cutoffs and 20 different k-mer sizes we showed that i) none of the genes could be assembled across all of the parameter space ii) assembly of each gene required an optimal set of parameter values and iii) these parameter values could be explained in part by different gene expression levels and different degrees of similarity between genes.ConclusionsTo obtain optimal transcriptome assemblies for allopolyploid plants, k-mer size and k-mer coverage need to be considered simultaneously across a broad parameter space. This is important for assembling a maximum number of full length ESTs and for avoiding chimeric assemblies of homeologous and paralogous gene copies.

Transcriptomics of salinity tolerance capacity in Arctic charr (Salvelinus alpinus): a comparison of gene expression profiles between divergent QTL genotypes

“…The final dataset used for de novo transcriptome assembly was 38.9% of its original size. The number of contigs in each assembly, spanning k-mers 33 to 89, was inversely related to k-mer size (Table 2), which is consistent with previous observations (17,18,21,28). The optimum k-mer sizes appeared to be 57 and 65, which had the highest values for most metrics typically used to assess assembly quality (e.g., mean and median contig lengths, N50, and N90), maximum contig length the exception.…”

Insights into the Evolution of Hydroxyproline-Rich Glycoproteins from 1000 Plant Transcriptomes

“…Removing redundancies from the 1KP sequence data was necessary because, like other short-read sequence data, these offers a wealth of information, but redundancy can occur from multiple sources, including splice variants, sequencing errors, recent duplications, chimeric transcripts, and polyploidy (Gruenheit et al, 2012;Yang and Smith, 2013;Xie et al, 2014). Sequence redundancies were obvious in our analysis (Supplemental Fig.…”

De Novo Assembly and Characterization of the Transcriptome of the Parasitic Weed Dodder Identifies Genes Associated with Plant Parasitism

“…Initially, the assembled transcriptome was highly enriched for small-sized transcripts, similar to other de novoassembled plant transcriptomes, which could in part be due to the transcript assembly algorithm, where the reads are decomposed into k-mers that may report hypothetical and misassembled transcripts (Zhao et al, 2011;Gruenheit et al, 2012). Filtering out transcripts based on abundance estimation and clustering by sequence identity would have likely removed these assembly artifacts, including transcript variants that were poorly supported by the reads.…”