Genome-Wide Association Scans and Identification of Candidate Genes

The association mapping panel was genotyped using the 9k iSelect SNP array (Illumina, CA, United States). SNPs were filtered for missing data (5%) and minor allele frequency (MAF < 0.05), resulting in 4,866 polymorphic SNPs (DOI: 10.5447/IPK/2019/14) further used for genetic analysis with 4,122 SNPs having mapping information as described in Neumann et al. (2017). The decay of linkage disequilibrium (LD) and the structure of the association mapping panel were estimated by Neumann et al. (2017). The average LD decay amounted to 8 cM in the collection and a kinship using modified Rogers’ Distance (Reif et al., 2005) was sufficient to correct for population structure and included in the model for genome-wide association scans (GWAS). GWAS were performed using BLUEs from single experiments using the software ASReml-R 3.0 (Butler et al., 2009). The following mixed-linear model was applied as in Neumann et al. (2017):

where μ is the overall mean, and E is the effect of the experiments, S is the effect of SNP, and G is the random effect of the genotype, while e is the residual error. This model considers a covariance structure of 2K σG2 , where K refers to the kinship matrix (Jiang et al., 2015) and σG2 is the genetic variance. A false discovery rate (FDR) with a significance level of 0.1 was applied for each trait and separately for all days (Benjamini and Hochberg, 1995). The proportion of explained genetic variance (GV) of the detected QTL was estimated as the adjusted r² values standardized with the heritability.

We explored all loci for potential candidate genes using the recently annotated barley genome assembly (Mascher et al., 2017) using the gene sets from 2012 for genetic positions and 2016 for physical map positions (Colmsee et al., 2015). Designation of the HvARF genes was performed according to Tombuloglu (2019).