Monday, August 18, 2008

Asia finally getting in on the genome-wide association game?   posted by p-ter @ 8/18/2008 07:25:00 PM

Nature Genetics has reports from two Japanese groups on an association between variants in KCNQ1 and susceptibility to Type II diabetes. The study itself falls into the now-standard genome-wide association study mold (except perhaps for the choice of SNPs in one of the studies--"only" 100,000, mostly genic--and the decision to not type them using a chip), but performing GWA studies in non-European populations is important for a number of reasons:

1. It enables the identification of novel loci involved in a trait. In some idealized model, risk for a disease could hypothetically be influenced by a set of N genes, but in a population polymorphic at only some subset of those genes, only that subset will be found by association studies (which map variation--if there's no variation, an association study won't find anything). Due to the vagaries of genetic drift (and perhaps natural selection), different populations may end up polymorphic at a different subset of those genes. So mapping the same trait in a number of distant populations could lead to a much greater understanding of the biology of the trait.

Even if it's not the case that one population has variation in a gene and another population doesn't, it's certainly the case that the power to detect a given variant can differ between populations due to smaller variations in allele frequency. That seems to be the case here--the SNP identified is at low frequency in European populations (~5% minor allele frequency), but at modest frequency in East Asians (~40% MAF), right in the "sweet spot" for detection by an association study.

2. Studies of the same phenotype in multiple populations also allows one to get at questions of "heterogeneity". Back in the old days of candidate gene approaches to complex disease, the failure to replicate an association was often blamed on heterogeneity, ie. the possibility that the risk alleles in your population were different from the risk alleles in the population in which the association was originally found. That excuse always seemed to be a bit suspect, and now, with sample sizes orders of magnitude greater, it's testable. In this case, the risk allele in the Japanese sample is replicated in a Danish populations, suggesting heterogeneity (and effects of genetic background) are not an issue in this instance.