Journal of Heredity Advance Access published online on April 1, 2009
Journal of Heredity, doi:10.1093/jhered/esp012
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
Intrabreed Stratification Related to Divergent Selection Regimes in Purebred Dogs May Affect the Interpretation of Genetic Association Studies
From the Department of Psychiatry and Institute for Human Genetics, University of California, San Francisco (Chang, Yokoyama, Hamilton); the Center for Neurobiology and Behavior, Department of Psychiatry, University of Pennsylvania, Philadelphia (Branson, Dyer, Overall); and the Institut de Génétique et Développement, CNRS UMR6061, Université de Rennes, 2 Av du Pr. Léon Bernard, 35043 Rennes, France (Hitte)
Address correspondence to Steven P. Hamilton at the address above, or e-mail: steveh{at}lppi.ucsf.edu.
Until recently, canine genetic research has not focused on population structure within breeds, which may confound the results of case–control studies by introducing spurious correlations between phenotype and genotype that reflect population history. Intrabreed structure may exist when geographical origin or divergent selection regimes influence the choices of potential mates for breeding dogs. We present evidence for intrabreed stratification from a genome-wide marker survey in a sample of unrelated dogs. We genotyped 76 Border Collies, 49 Australian Shepherds, 17 German Shepherd Dogs, and 17 Portuguese Water Dogs for our primary analyses using Affymetrix Canine v2.0 single-nucleotide polymorphism (SNP) arrays. Subsets of autosomal markers were examined using clustering algorithms to facilitate assignment of individuals to populations and estimation of the number of populations represented in the sample. SNPs passing stringent quality control filters were employed for explicitly phylogenetic analyses reconstructing relationships between individuals using maximum parsimony and Bayesian methods. We used simulation studies to explore the possible effects of intrabreed stratification on genome-wide association studies. These analyses demonstrate significant stratification in at least one of our primary breeds of interest, the Border Collie. Demographic and pedigree data suggest that this population substructure may result from geographic isolation or divergent selection regimes practiced by breeders with different breeding program goals. Simulation studies indicate that such stratification could result in false discovery rates significant enough to confound genome-wide association analyses. Intrabreed stratification should be accounted for when designing and interpreting the results of case–control association studies using purebred dogs.
Key Words: Bayesian analysis • canine genetics • maximum parsimony • phylogenetics • population stratification • purebred dogs
Corresponding Editor: Francis Galibert
Received November 23, 2008
Revised February 13, 2009
Accepted February 25, 2009