Repeatability of Genotyping By Sequencing SNP Markers in Wheat: An Investigation

Genotyping by sequencing (GBS) provides a snapshot view of the genome and generates SNP markers that are free of ascertainment bias at a low cost. However, this approach suffers from data sparsity and high proportion of missing alleles upon multiplexing of large number of samples leading to low repeatability of genetic markers. Our investigation behind the lack of connectedness among GBS markers in two RIL populations LMPG-6/Paka and LMPG-6/Romany showed that only 1185 SNP markers (9.9% of 11998 total markers) were common. However, a search for common markers within the same recombination bins in these two populations led to detection of over 5000 common markers. We also found that using enzymes with higher cut frequency leads to a large amount of DNA fragments, which in turn increases the sampling of non-overlapping reads from different parts of the genome. This appears to be the main reason why not all genomic regions have uniform read coverage upon sequencing. We also investigated to what extent dense and sparse multiplexing levels exacerbate marker repeatability by simulating GBS library preparation and sequencing. As expected, multiplexing in higher order decreased the number of SNPs in 192-plex and 384-plex libraries relative to 48- and 96-plexed libraries. Read depth per SNP site reduced by a factor of ½ each time to 2.5, 1.2, and 0.6 in 96-plexed, 192-plexed, and 348-plexed panels, respectively, from that of 5 in 48-plexed library. For best results, sequencing 92-plex libraries appears to give relatively good genome coverage while still keeping the cost low.