Analysis of the effect of genetic heterogeneity on de novo genome assembly using Xylocopa virginica as a model.
Next generation sequencing (NGS) technology has revolutionized genomic and genetic research, and as a result, de novo genome sequencing and assembly for non-model organisms has now become a common task in genome research. However, the integral properties of a genome such as ploidy, mutations, and repeat content impose issues for current genome assemblers. In this work, we used Xylocopa virginica (Eastern carpenter bees) as a unique model organism for examining on the effect of sequence heterozygosity on quality of de novo genome assembly. Using two de Bruijn graph genome assemblers, we assembled four bee genomes representing different sex and age (unworn male, worn male, unworn female, worn female) using standard Illumina sequencing and one genome using 10X linked-reads library for an unworn female. We discovered that there is a noticeable difference in a variety of genome assembly quality metrics, with the haploid unworn male genome having the highest quality and the worn diploid female genome having the lowest quality. In fact, the N50 value of the unworn male genome was >100 times higher than that of the worn female genome. The genome quality pattern supports the hypothesis that sequence heterozygosity resulting both from ploidy and somatic variants can affect the result of an assembly with former shown to be a much bigger player than the latter. Furthermore, we observed that the density of variants was moderately correlated to the density of breakpoints in the genome assemblies. Overall, our results indicate that increased ploidy and accumulation of somatic variants both negatively affect the quality of the resulting assembly with the former being much more significant than the latter. When considering a de novo assembly project for a non-model organism, whenever possible, haploid samples at the youngest possible age are to be recommended. Furthermore, use of a long-read platform can lead to better genome quality. However, at least for the 10x linked reads, having too much sequencing data does not necessarily lead to a better genome assembly.