Figure 3

Local haplotype inference. (A) A three-SNP window is shown across 25 strains. A "0" denotes a minor allele (at the bi-allelic locus) and "1" the major allele. An "N" denotes no information for that strain, and strains lacking genotype data at any SNP position in a given window are not included in association computations at that locus. The three-SNP window shown contains three inferred haplotype groups with sizes of nine, twelve, and four strains per group. Groups with fewer than three strains are not considered in the analysis. This three-SNP window, which overlaps a previously known QTL, illustrates why the genotype data at a single locus is insufficient to model the genomic structure across inbred strains. (B) A histogram showing the number of haplotype groups for window sizes of two SNPs (black), three SNPs (red) four SNPs (blue) and five SNPs (green) is shown. Inferred haplotype groups with less than three members are excluded from the calculations and are not counted here. While two-SNP windows often show the maximum of four haplotype groups, larger SNP window sizes rarely exceed eight haplotype groups. This observation suggests that three-SNP windows are most appropriate for this strain set and SNP data. (C) A three-SNP window is shown with highly unequal group sizes and unequal variances. This locus is identified as significant by IH-KW but not IH-P.