Genotyping-By-Sequencing of a Diploid Potato F2 Population

Genotyping-by-sequencing (GBS) of multiplexed, restriction-site associated DNA (RAD) libraries is an attractive technology for generating genome-wide markers because of its technical simplicity and low costs per sample.  To investigate its feasibility for potato, a diploid F2 population (S. tuberosum DM1-3 x S. chacoense M6) was genotyped at 96-plex using the enzyme EcoT22I.  Using an R-based bioinformatics pipeline to filter the GBS variants, more than 11,000 bi-allelic SNPs were identified.  During the curation process, we identified variants in 45 superscaffolds whose linkage group conflicted with the chromosome assignment in the latest version (4.03) of the potato reference genome; further study of these discrepancies may help improve the reference sequence.  One indicator of the quality of the curated marker data was that the phase of the F1 parent, which was predicted from the F2 progeny using a Hidden Markov Model, was correct across the entire genome.  Another indicator of quality was our ability to map two traits—pigmented skin color (Stan2) and yellow flesh color (CHY2)—to their known positions on chromosomes 10 and 3, respectively.  To guide future research, we used the empirical read depth distribution to simulate GBS libraries at several multiplex levels.  For diploid potato, the 96-plex level used in this study appears to offer the best compromise between marker density and cost per sample.  For autotetraploid potato, because greater depth is needed to differentiate the three heterozygous genotypes, 48-plex (or lower) is recommended when using EcoT22I.