Effects of population stratification on inferences of marker effects and Association tests

Genome Wide Association Studies (GWAS) have reported unprecedented numbers of variants associated to important human traits and diseases. This information can be used to uncover genetic factors leading to important medical conditions. Commonly, the association between markers (Single Nucleotide Polymorphisms) and phenotypes is assessed using Single Marker Regression (SMR). However, the effects of individual variants on genetic risk may be affected by Population Stratification (PS). A commonly used approach to account for PS consists of expanding the SMR with inclusion of Principal Components (PC); this approach accounts for possible additive effects of PCs. Using a collection of wheat lines which are known to exhibit great degree of PS, we evaluated the extent to which the effects of markers may be modulated. Our baseline model ignores the effects of PS, in a second model PS was accounted for by including additive effects of the first 2 PC’s. Finally, we considered a third model which accounts for additive and interaction effects of markers and PC’s. Inclusion of the additive effects of PC in Model 2 had little impact on inferences. We found more markers exhibiting significant interaction (94, out of 1,279 markers tested in Model 3) than markers exhibiting significant main effects (24, out of 1,279 main effects tested in Model 3). A percentage (29%) of markers exhibiting significant interactions in Model 3 did not exhibit significant main effect in Model 2. In summary, there was substantial evidence that additive effects of markers are modulated by genetic background.