Linkage disequilibrium (LD) estimates are often calculated genome-wide for use in many tasks, such as SNP pruning and LD decay estimation. However, in the presence of genotype uncertainty, naive approaches to calculating LD have extreme attenuation biases, incorrectly suggesting that SNPs are less dependent than in reality. These biases are particularly strong in polyploid organisms, which often exhibit greater levels of genotype uncertainty than diploids. A principled approach using maximum likelihood estimation with genotype likelihoods can reduce this bias, but is prohibitively slow for genome-wide applications. Here, we present scalable moment-based adjustments to LD estimates based on the marginal posterior distributions of the genotypes. We demonstrate, on both simulated and real data, that these moment-based estimators are as accurate as maximum likelihood estimators, but are almost as fast as naive approaches based only on posterior mean genotypes. This opens up bias-corrected LD estimation to genome-wide applications. In addition, we provide standard errors for these moment-based estimators. All methods discussed in this manuscript are implemented in the ldsep package, available on the Comprehensive R Archive Network (https://cran.r-project.org/package=ldsep).

连锁不平衡 (LD) 估计通常在全基因组范围内计算，用于许多任务，例如 SNP 修剪和 LD 衰减估计。然而，在存在基因型不确定性的情况下，计算 LD 的简单方法存在极大的衰减偏差，错误地表明 SNP 的依赖性低于实际情况。这些偏差在多倍体生物中尤其强烈，多倍体生物通常表现出比二倍体更高水平的基因型不确定性。使用最大似然估计和基因型可能性的原则性方法可以减少这种偏差，但对于全基因组应用来说速度太慢。在这里，我们根据基因型的边际后验分布对 LD 估计进行可扩展的基于矩的调整。我们在模拟和真实数据上证明，这些基于矩的估计器与最大似然估计器一样准确，但几乎与仅基于后验平均基因型的朴素方法一样快。这为全基因组应用开辟了偏差校正 LD 估计。此外，我们还为这些基于矩的估计器提供标准误差。本手稿中讨论的所有方法均在ldsep包中实现，可在综合 R 存档网络 (https://cran.r-project.org/package=ldsep) 上找到。