CSS for complex traits
The CSS approach demonstrates that the test is generally applicable for the identification of trait-specific genomic regions for complex traits by comparing phenotypically contrasting populations. The CSS test combines multiple pieces of evidence from the rank distribution of different selection tests in a weighted index of signatures of selection. The power of the test and its constituents (FST, ΔDAF, and XP-EHH) depends on the phenotypic and genetic divergence between the candidate and reference populations, as we demonstrated with simple binary monogenic (Randhawa et al. 2014) and complex polygenic traits (this study).
Use of multibreed cohorts has increased the discovery of trait-specific regions because of shared linkage disequilibrium between the causal mutations and neighboring SNPs (Kemper and Goddard 2012) arising from long-term historical selection. Contrasting patterns of genomic variation at the putative regions under selection across groups increases the likelihood of capturing the signatures of selection linked to stature. In our approach, signals of selection are amplified at candidate regions across breeds within groups, whereas background noise (false-positive signals) is reduced in the rest of the genome. Such noise is usually expected from the demographic history of breed formation and random genetic drift (Seichter et al. 2012). Overall, application of the CSS method combined with grouping breeds according to their wither height identified candidate regions in the bovine genome for this complex trait.
Detection of signatures of selection is valuable in the discovery of potential genomic regions of functional mutations affecting quantitative traits (Hayes et al. 2008). Recent investigations suggest that, in the absence of classic selective sweeps, signatures of selection for complex traits are less likely to be detected by using individual selection tests on single breed data (Kemper et al. 2014). Our approach combines multiple selection tests and grouping of phenotypically alike breeds, and has been found to be highly efficient at detecting signatures of selection and identifying candidate gene regions for complex traits. This approach makes use of existing resources under long-term historical selection and provides a relatively inexpensive entry for more detailed follow-up studies in the genetic architecture of complex traits.