Case-Control Association Tests Correcting for Population Stratification

Abstract

A widely discussed problem in genetic case-control association studies is the impact of population stratification when testing for an association between a genetic marker and the disease. Genotyping of additional genetic markers is necessary to eliminate spurious associations caused by population stratification. Association tests accounting for population structure broadly follow one of two concepts: Genomic Control or Structured Association. Genomic Control tests attempt to empirically estimate the variance inflation of the original test statistic. Here we theoretically investigate Genomic Control and present some new results about the type-I error rate of Genomic Control. The main focus of this thesis is on the Structured Association approach where population structure is directly inferred and the test of association incorporates the estimated population structure. To model population structure we assume that the total population is a mixture of discrete subpopulations. We show that it is necessary to incorporate phenotypic information when inferring population structure, otherwise a systematic bias is introduced. Moreover, for moderate population stratification we derive a new Wald test statistic. To assess the impact of population stratification on case-control studies within Germany, data of the German Genomic Control Study are analyzed. In simulations we compare the introduced extensions of Structured Association to existing methods of Structured Association as well as to Genomic Control. Realistic situations of large case-control studies with moderate population stratification are considered. The simulations show that the model based approach of Structured Association, if applied appropriately, is in general superior to the Genomic Control approach, at least in the case of simple population structure as simulated here.