Extended Data Fig. 9: GWAS Manhattan plots exploring sensitivity of QTL detection for reduced resolution genotype data, and simulated loci varying in effect size and frequency.

Manhattan plot showing impact of marker density on discovery of non-additive bodyweight GWAS signals (a; P-values computed using Z-tests, horizontal grey line indicates the genome-wide significance threshold of P < 5 × 10⁻⁸). Here, dominance estimates from sequence-based bodyweight GWAS (grey dots) are plotted alongside a subsetted version of these same data filtered to represent the content of BovineSNP50k SNP-chip platform (green dots). While two of the modest effect, comparatively higher MAF QTL retain significance (i.e. Chr2:22Mbp and PLAG1 locus), only the DPF2/MUS81 QTL is represented among the major-effect, recessive signals. (b) Manhattan plot showing the influence of MAF and effect size on sensitivity of detection in a simulated dataset. Dominance estimates (blue dots) are contrasted with standard-additive estimates (grey dots), showing sensitivity of detection for 30 recessive causative mutations (red dots). Recessive effects were generated by randomly selecting variants from 1-5% MAF bins from the pool of simulated genotypes (frequencies indicated at bottom), with effect sizes assigned as 0.5 standard deviations (SD; light orange) or 1.0 SD (dark orange) per mutation. Mutations were selected to represent all chromosomes (two on chromosome 1).