Mouse SNPs for evolutionary biology: Beware of ascertainment biases

Recently, Harr (2006) used publicly available Single Nucleotide Polymorphism (SNP) data (http://www.well.ox.ac.uk/mouse/INBREDS/) to search the mouse genome for regions of elevated differentiation between the subspecies of the house mouse. If they existed, such regions would be of great interest because they would be likely to harbor the genetic factors causing the partial incompatibility between the subspecies, a phenomenon particularly documented in the case of the European subspecies Mus musculus domesticus and Mus musculus musculus (Storchova et al. 2004; Britton Davidian et al. 2005; Raufaste et al. 2005). Using the 10,265 SNPs spanning the whole genome and that have been typed in seven domesticus and eight musculus wild-derived strains of various geographical origins, Harr reports that 10 regions located on seven different chromosomes show a significantly higher than average regional proportion of SNPs with alleles alternatively fixed in the two subspecies.

Using SNP data in population genetics and evolution, especially for quantifying differentiation or divergence, requires a careful and unbiased choice of SNPs. Ascertainment biases in the SNP discovery and choice processes can have various origins and can sometimes lead to erroneous conclusions (Morin et al. 2004; Clark et al. 2005). Most mouse SNPs have been discovered by comparing the genomes of classical inbred laboratory mouse strains that have long been known (Bishop et al. 1985) to be hybrids between M. m. domesticus (the Western European subspecies) and mice of Asian origin (M. m. musculus and/or Mus musculus castaneus). Recent surveys have …