Separating homeologs by phasing in the tetraploid wheat transcriptome

KRASILEVA KV; BUFFALO V; BAILEY P; PEARCE S; TABBITA F; SORIA MA; WANG S; THE IWGS CONSORTIUM; AKHUNOV E; UAUY C; DUBCOVSKY J

GENOME BIOLOGY

BIOMED CENTRAL LTD

Lugar: Londres; Año: 2013 vol. 14 p. 1 - 43

1474-760X

BackgroundThe high level of identity among duplicated homoeologous genomes in tetraploid pasta wheatpresents substantial challenges for de novo transcriptome assembly. To solve this problem, wedevelop a specialized bioinformatics workflow that optimizes transcriptome assembly andseparation of merged homoeologs. To evaluate our strategy, we sequence and assemble thetranscriptome of one of the diploid ancestors of pasta wheat, and compare both assemblies with abenchmark set of 13,472 full-length, non-redundant bread wheat cDNAs.ResultsA total of 489 million 100 bp paired-end reads from tetraploid wheat assemble in 140,118contigs, including 96% of the benchmark cDNAs. We used a comparative genomics approach toannotate 66,633 open reading frames. The multiple k-mer assembly strategy increases theproportion of cDNAs assembled full-length in a single contig by 22% relative to the best singlek-mer size. Homoeologs are separated using a post-assembly pipeline that includespolymorphism identification, phasing of SNPs, read sorting, and re-assembly of phased reads.Using a reference set of genes, we determine that 98.7% of SNPs analyzed are correctlyseparated by phasing.ConclusionsOur study shows that de novo transcriptome assembly of tetraploid wheat benefit from multiplek-mer assembly strategies more than diploid wheat. Our results also demonstrate that phasingapproaches originally designed for heterozygous diploid organisms can be used to separate theclose homoeologous genomes of tetraploid wheat. The predicted tetraploid wheat proteome andgene models provide a valuable tool for the wheat research community and for those interestedin comparative genomic studies.