Investigation of Chimeric Gene Assembly in a Polyploid Genome

Heterozygous diploid and polyploid genomes present two or more alternative variants for assembly programs.  The correct assembly of variants depends upon the scale and frequency of identical regions among the variants, as well as sequencing depth, error frequency, and number of variants.  When a conserved subsequence is longer than its matching reads or mate pairs, correct linking of non-identical flanking subsequences cannot be guaranteed.  To elucidate the effects of subsequence structure, sequencing depth, error frequency, and number of similar sequences, simulated reads were generated with mean length and standard deviation similar to the 454 reads reported by Brenchley et al. (2012) for hexaploid breadwheat.  Assembly was simulated of an 1800-base relatively conserved region flanked by 1800-base less-conserved regions, using an iterative matching-retrieval-assembly pipeline consisting of blastn, blastdbcmd, and phrap.  Besides correct assemblies, block and scrambled chimeric assemblies resulted.  Block chimeras had SNPs in two to four large, internally correct regions.  Scrambled chimeras had SNPs drawn almost randomly from multiple variants.  Conserved regions often were merged into too few contigs, which isolated the extra flanking subsequences as distinct contigs.  Contigs were termed “complete” if they spanned the central region and all or most of both flanking regions, and “incomplete” otherwise.  Increased sequencing depth from 5x to 10x increased the fraction of complete contigs, but increased sequencing error and ploidy had little effect.  As expected, increased ploidy increased the frequency of block chimeras among complete contigs, and increased sequencing error somewhat increased the frequency of scrambled chimeras.