Assessing the Assembly Quality of Large Plant Genomes

Large amounts of repetitive DNA hampers de novo assembly of large plant genomes, particularly when assembling short reads with current computational methods. As part of the IWGSC, we assembled the hexaploid wheat genome from Illumina shotgun data of individual flow-sorted chromosome arms. In this study, we have reassembled four out of 41 chromosome arms and used a mixture of traditional and new kmer-based metrics on the read data and assemblies to discover the optimal parameters for the assembly process.

To check quality and information content of the read data, we looked at the kmer frequency distribution before and after quality score based trimming. As expected, trimming discarded both errors and correct data, but we were unable to see any improvement in the assemblies based on this kmer analysis and on the traditional metrics that were based on length and coverage.

Upon assessing assembly quality with different assemblers and parameters, the traditional metrics were unable to reveal the best assembly for each chromosome arm, so we explored kmer space based approaches following the same principles we applied to read checking. We performed a kmer frequency comparison between the read data and the assembly to assess how well the assemblies represent the read data. We will show how a combination of these new metrics with the more traditional length and coverage based metrics has enabled us to identify the best assembly for each chromosome arm.