The effects of restrictionenzyme choice on properties of genotypingbysequencing libraries: a study in cassava (Manihot esculenta)

Abstract

Compared with other reduced-representation sequencing methods, library construction for genotyping-by-sequencing (GBS) is simpler and less expensive. However, elimination of sizeselection steps results in libraries of more variable fragment size than with other reduced-representation methods, affecting several aspects of the data. To test the effect of restriction enzyme choice on library quality, we made GBS libraries with PstI (6-cutter), PstI/TaqI (4-cutter), or ApeKI (4.5-cutter) from the same set of DNAs from a cassava (Manihot esculenta Crantz) biparental population. Tag and single nucleotide polymorphism (SNP) counts were limited by the number of cut sites rather than by read number. Depth per locus was very skewed for the PstI library, such that most SNPs had low read depth but a subset had very high read depth. In contrast, the ApeKI and PstI/TaqI libraries had less variable distributions of read depth and yielded far more scorable SNPs. Our results suggest that 6-cutter enzymes may be most appropriate for genotyping a modest number of markers at a high multiplexing level, or for very large genomes, and perform better when used in a double digest with a 4-cutter enzyme.