Local GBS Haplotypes in Hexaploid Oat: How and Why?

Genotyping-by-sequencing (GBS) is a cost-effective high-throughput genotyping system.  However, analysis of GBS data is challenging, especially in a polyploid species with no reference genome (e.g. hexaploid oat) where intra- and inter-locus polymorphism are easily confounded.  Pipelines that call SNPs from de-novo assemblies of multiple sequence-tags are prone to error when the tags originate from multiple loci.  This problem is addressed by excluding SNPs that show excessive heterozygosity or segregation distortion in populations where they shouldn’t.  The UNEAK pipeline further simplifies GBS analysis by restricting SNP-calls to pairs of GBS-tags that contain a single reciprocal SNP.  However, many useful polymorphisms from tags that contain multiple SNPs are excluded by these methods.  In an effort to address this, we developed a novel experimental pipeline that considers all tags from a given assembly as potential haplotypes from a single locus complex.  We then test all possible subsets of haplotypes within the assembly to identify the ‘best’ model: fitting the largest number of genotypes while satisfying population-based frequency thresholds.  This pipeline initially identifies what we call “local haplotypes”, analogous to multiple alleles, which are directly useful in population-based genetic analyses.  We have applied this pipeline to a previously analysed set of data from 2665 accessions originating from mixed populations in several oat species.  Our results demonstrate several potential benefits, including: direct assignment of locus homology between populations and species; haplotype-based GWAS, and accurate identification of consensus and reference sequences to assist in genome assembly. Additionally, we provide an intuitive graphical 'passport' for each haplotype model.