Methods to detect polygenic adaptation have recently been shown to be sensitive to uncorrected stratification in GWAS, thereby casting doubts on whether polygenic adaptation is prevalent among humans. Consistent with a signal of adaptation at human height loci, the mean F_ST among African, East Asian, and European populations was shown to be significantly higher at height-associated SNPs than that at non-associated SNPs. This conclusion was reached, however, using height-associated SNPs ascertained from a GWAS design impacted by residual confounding due to uncorrected stratification. Specifically, we show here that the estimated effect sizes are significantly correlated with population structure across continents, potentially explaining the elevated differentiation previously reported. We alleviated these concerns of confounding by ascertaining height-associated SNPs from two biobank GWAS (UK Biobank, UKB, and Biobank Japan, BBJ), where measures to control for confounding in GWAS are more effective. Consistent with a global signature of polygenic adaptation, we found that compared to non-associated SNPs, frequencies of height-associated SNPs are indeed significantly more differentiated among continental populations from both the 1000 Genomes Project (p = 0.0012 for UKB and p = 0.0265 for BBJ), and the Human Genome Diversity Project (p = 0.0225 for UKB and p = 0.0032 for BBJ). However, we found no significant difference among continental populations in polygenic height scores. Through simulations, we found that polygenic score-based statistics could lose power in detecting polygenic adaptation in presence of independent converging selections, thereby potentially explaining the inconsistent results based on F_ST and polygenic scores.

检测多基因适应的方法最近已被证明对 GWAS 中未校正的分层敏感，从而使人们怀疑多基因适应是否在人类中普遍存在。与人类身高位点的适应信号一致，平均 F _ST在非洲、东亚和欧洲人群中，与身高相关的 SNP 显着高于非相关 SNP。然而，这个结论是使用从 GWAS 设计中确定的高度相关 SNP 得出的​​，该设计受到由于未校正的分层导致的残余混杂影响的影响。具体来说，我们在这里表明，估计的效应大小与各大洲的人口结构显着相关，这可能解释了先前报道的高分化。我们通过从两个生物库 GWAS（UK Biobank, UKB 和 Biobank Japan, BBJ）中确定与身高相关的 SNP 来缓解这些混杂问题，其中控制 GWAS 混杂的措施更有效。与多基因适应的全球特征一致，我们发现与非相关 SNP 相比，UKB 为p  = 0.0012， BBJ 为p  = 0.0265），以及人类基因组多样性计划（ UKB 为p  = 0.0225， BBJ 为p  = 0.0032）。然而，我们发现大陆人群的多基因身高评分没有显着差异。通过模拟，我们发现基于多基因评分的统计数据在存在独立收敛选择的情况下可能会失去检测多基因适应的能力，从而可能解释基于 F _ST和多基因评分的不一致结果。