Singh Archit, Southam Lorraine, Hatzikotoulas Konstantinos, Rayner Nigel W, Suzuki Ken, Taylor Henry J, Yin Xianyong, Mandla Ravi, Huerta-Chagoya Alicia, Morris Andrew P, Zeggini Eleftheria, Bocher Ozvan
Technical University of Munich (TUM), TUM School of Medicine and Health, Graduate School of Experimental Medicine, Munich, Germany.
Institute of Translational Genomics, Helmholtz Zentrum München- German Research Center for Environmental Health, Neuherberg, Germany.
Genet Epidemiol. 2025 Sep;49(6):e70016. doi: 10.1002/gepi.70016.
Inflation in genome-wide association studies (GWAS) summary statistics represents a major challenge, for which correction methods have been developed. These include the genomic control (GC) method, which uses the λ-value to correct summary statistics, and the linkage disequilibrium score regression (LDSR) method, which uses the LDSR intercept. By using type 2 diabetes (T2D) as an exemplar, we explore factors influencing λ-values and the impact of these corrections on association signals. We find that larger sample sizes increase λ-values due to increased captured polygenicity, while including lower frequency variants decreases λ-values due to reduced power. Comparing T2D genetic associations described in overlapping GWAS meta-analyses of increasing sample size, we find that GC correction reduces the false positive rate and leads to the loss of robust associations. In one of the largest meta-analysis, GC correction results in 39.7% loss of independent loci, substantially reducing the number of detected associations. In comparison, the LDSR intercept correction leads to a loss of up to 25.2% of the independent loci, being therefore less conservative than the GC correction. We conclude that in large, well-powered GWAS meta-analysis of polygenic traits, both GC and LDSR intercept correction leads to power loss, highlighting the need for improved genomic inflation correction methods.
全基因组关联研究（GWAS）汇总统计中的膨胀是一个重大挑战，针对此已开发出校正方法。这些方法包括基因组控制（GC）方法，该方法使用λ值校正汇总统计，以及连锁不平衡评分回归（LDSR）方法，该方法使用LDSR截距。以2型糖尿病（T2D）为例，我们探讨影响λ值的因素以及这些校正对关联信号的影响。我们发现，更大的样本量会因捕获的多基因性增加而使λ值升高，而纳入低频变异会因效能降低而使λ值降低。比较不同样本量的重叠GWAS荟萃分析中描述的T2D遗传关联，我们发现GC校正降低了假阳性率，但导致了稳健关联的丢失。在一项最大的荟萃分析中，GC校正导致39.7%的独立位点丢失，大幅减少了检测到的关联数量。相比之下，LDSR截距校正导致高达25.2%的独立位点丢失，因此不如GC校正保守。我们得出结论，在多基因性状的大型、高效能GWAS荟萃分析中，GC和LDSR截距校正都会导致效能损失，这凸显了改进基因组膨胀校正方法的必要性。
