Shin Jimin, Lee Chaeyoung
Department of Bioinformatics and Life Science, Soongsil University, Seoul 156-743, Republic of Korea.
Department of Bioinformatics and Life Science, Soongsil University, Seoul 156-743, Republic of Korea.
Genomics. 2015 Apr;105(4):191-6. doi: 10.1016/j.ygeno.2015.01.006. Epub 2015 Jan 30.
Population stratification can produce spurious genetic associations in genome-wide association studies (GWASs). Mixed model methodology has been regarded useful for correcting population stratification. This study explored statistical power and false discovery rate (FDR) with the data simulated for dichotomous traits. Empirical FDRs and powers were estimated using fixed models with and without genomic control and using mixed models with and without reflecting loci linked to the candidate marker in genetic relationships. Population stratification with admixture degree ranged from 1% to 10% resulted in inflated FDRs from the fixed model analysis without genomic control and decreased power from the fixed model analysis with genomic control (P<0.05). Meanwhile, population stratification could not change FDR and power estimates from the mixed model analyses (P>0.05). We suggest that the mixed model methodology was useful to reduce spurious genetic associations produced by population stratification in GWAS, even with a high degree of admixture (10%).
群体分层会在全基因组关联研究（GWAS）中产生虚假的基因关联。混合模型方法被认为有助于校正群体分层。本研究利用为二分性状模拟的数据探究了统计功效和错误发现率（FDR）。使用有无基因组控制的固定模型以及在遗传关系中有无反映与候选标记连锁位点的混合模型来估计经验性FDR和功效。混合程度在1%至10%之间的群体分层导致无基因组控制的固定模型分析中FDR膨胀，以及有基因组控制的固定模型分析中功效降低（P<0.05）。同时，群体分层不会改变混合模型分析中的FDR和功效估计值（P>0.05）。我们认为，即使在高度混合（10%）的情况下，混合模型方法对于减少GWAS中群体分层产生的虚假基因关联也是有用的。
