Fig 1.
Simulated allele frequency trajectories and model overview.
(a-c) Allele frequency trajectories for 20 SNPs colored by relative effect sizes from stochastic selection simulations. (a) Effect size = 0, representing stochastic changes in allele frequency due to genetic drift. (b) Large-effect alleles rapidly becoming fixed in the population representing selective sweeps. (c) Moderate-to-small effect size SNPs changing in frequency slowly over time, representing polygenic selection. (d) An overview of the linear mixed model approach used for Generation Proxy Selection Mapping and environmental GWAS. (e-f) A single SNP under ecoregion-specific selection. Different colors represent the trajectory of a given SNP in one of five different ecoregions. Ecoregion-specific selection can lead to allele frequencies that (e) diverge from or (f) converge to the population mean. Maps were plotted using public domain data from the US Department of Commerce, Census Bureau via the R package maps (version 3.1, https://cran.r-project.org/web/packages/maps/).
Fig 2.
Generation Proxy Selection Mapping identifies signals of polygenic selection in three major U.S. cattle populations.
Full and truncated (-log10(q)< 15) Manhattan plots for GPSM analysis of Red Angus (a & b), Simmental (c & d), and Gelbvieh (e & f). Purple points indicate SNPs significant in all three population-specific GPSM analyses and orange points indicate SNPs significant in two. Red lines in a-f indicate a significance cutoff of q < 0.1. Quantile-quantile plots of -log10(p) values in GPSM analysis of (g) Red Angus, (h) Simmental, and (i) Gelbvieh populations. Minor allele frequency plotted versus -log10(p) values for significant SNPs in (j) Red Angus, (k) Simmental, and (l) Gelbvieh populations. (m) Smoothed allele frequency histories for the six most significant loci identified as being under selection in all three datasets. (n) Allele frequency histories for three known Mendelian loci that control differences in visual appearance between introduced European and modern US Simmental cattle. Arrows of the same color are used to distinguish the genomic locations of these loci in (c).
Table 1.
Number of significant birth date-associated SNPs in each population at various significance thresholds.
Table 2.
Variation in birth date explained by three classes of SNPs.
The PVE estimates (standard error in parentheses) from a genomic restricted maximum likelihood (GREML) variance component analysis of birth date using three GRMs created from: 1) genome-wide significant SNPs (q < 0.1), 2) an equivalent number of the next most significant SNPs outside of genome-wide significant associated regions, and 3) an equivalent number of randomly sampled SNPs from genomic regions that did not harbor genome-wide significant associations.
Fig 3.
Manhattan plots for discrete and continuous envGWAS in Red Angus cattle.
(a) Nine continental US ecoregions defined by K-means clustering of 30-year normal temperatures, precipitations, and elevations. (b) Locations of sampled Red Angus animals colored by breeder’s ecoregion and sized by the number of animals at that location. (c) Multivariate discrete envGWAS (case-control for six regions with > 600 animals). Locations of sampled Red Angus animals colored by (d) 30-year normal temperature, (e) 30-year normal precipitation, and (f) elevation. (g) Multivariate continuous envGWAS with temperature, precipitation, and elevation as dependent variables. For all Manhattan plots the red line indicates the empirically-derived p-value significance threshold from permutation analysis (p < 1⨯10−5). Maps were plotted using public domain data from the US Department of Commerce, Census Bureau via the R package maps (version 3.1, https://cran.r-project.org/web/packages/maps/).
Fig 4.
Meta-analysis of within-ecoregion GPSM for Red Angus cattle.
(a) Manhattan plot of per-variant Cochran’s Q p-values. Points colored green had significant Cochran’s Q (p < 1⨯10−5) and were significant in at least one within-region GPSM analysis (p < 1⨯10−5). (b) Ecoregion effect plots for lead SNPs from six loci from (a). Points are colored by ecoregion and are sized based on Cochran’s Q value. (c) Ecoregion-specific allele frequency histories for SNPs from (b), colored by ecoregion.