Evaluating and improving genome-wide association studies (GWAS) in mice
The well-established utility of the laboratory mouse for discovering disease-modulating genes may have recently been extended. In 2010, researchers from Rosetta Inpharmatics LLC, a wholly-owned subsidiary of Merck & Co., Inc., with the help of Karine Lux at The Jackson Laboratory, demonstrated that genome-wide association studies (GWAS), if used in combination with linkage analyses, can expedite the discovery of disease alleles in mice (Su et al. 2010).
Because GWAS have been so successful at revealing disease-modulating genes in human populations, researchers have sought to apply them to laboratory mouse populations. The higher mapping resolution possible with an association study compared to a linkage study could save a significant amount of time and money. Additionally, the increasing genotyping resources for the mouse, notably large data sets of readily available, high density and strain-specific single nucleotide polymorphic (SNP) markers, could further reduce the time and money needed to identify gene-phenotype associations in mice. Until recently, however, GWAS applications in inbred mouse populations have been unreliable. Because selective pressure (non-random mating) has been used to perpetuate laboratory mouse strains for decades, the genetic diversity in their ancestry is limited and many strains are required to detect significant associations. Additionally, many regions in the genomes of the laboratory mouse are highly correlated, resulting in spurious associations. Combined, these factors often confound mouse GWAS results, yielding many false associations and overlooked real ones.
To quantify the extent and nature of the association signals from a typical mouse GWAS, the research team performed a GWAS for more than 1,000 expression traits with known expression quantitative trait loci (eQTL) in 370 mice from 19 JAX® Mice strains (average of 10 mice/sex/strain). The strains are predisposed to a range of metabolic phenotypes and represent the distinct genealogies of inbred mice. All the mice were maintained and their liver tissues collected at our In vivo Pharmacology Services facility under the supervision of Karine Lux.
Not surprisingly, the team found that the statistical power of a 19-strain GWAS is low. Spurious associations are frequent, and the signals for many of them are orders of magnitude more significant than the signals for real ones. Often, markers with the most significant P values do not map to the location of the true eQTL. A combined GWAS and linkage analysis, however, yields significantly higher statistical power and mapping resolution than either technique alone. These results indicate that researchers must carefully plan the experimental design of a mouse GWAS, especially in mouse populations with significant potential for spurious associations.
Su WL, Sieberts SK, Kleinhanz RR, Lux K, Millstein J, Molony C, Schadt EE. 2010. Assessing the prospects of genome-wide association studies performed in inbred mice. Mamm Genome Feb 5. [Epub ahead of print]