Genotyping By Sequencing In Cassava, a Clonal, Outbred Crop.

We are using GBS to generate genotype information for a genomic selection project in cassava, which has an estimated genome size of 760 Mb and is highly heterozygous. Initially, we genotyped our training population with libraries made from PstI-digested DNA, but subsequently switched to using ApeKI libraries, which yield many more markers and much less skewed read depth per locus. To deal with the uncertainty of genotype calls when read depth is low, we score genotypes as likelihoods rather than as categorical values. These genotypes were used in various genomic selection cross-validation schemes, some based on large amounts of imputed data. We are currently investigating the best way to impute missing data. In a separate project to genotype an outbred cassava biparental population, we made GBS libraries with PstI, PstI/TaqI, or ApeKI for the same set of 95 DNAs. Each library produced a distinct profile of tag number and read depth. Choice of enzyme must be empirically determined for each taxon, but our results suggest that 6-cutter enzymes may be most appropriate for genotyping a modest number of markers at a high multiplexing level.