It is increasingly clear that most complex traits and diseases have highly polygenic architectures, with a large number of causal variants of small effect. In order to understand these polygenic architectures, it is of interest to infer parameters such as the heritability explained by SNPs and the number of variants with non-negligible effects on the trait. Both of these quantities have been estimated using accuracies of polygenic risk scores (see above), as a function of the P-value threshold used to constrain the set of markers employed75,76. Computing polygenic risk scores requires individual-level data in the validation cohort, implying that these methods are not strictly summary statistic based. Recent work has shown that the information in polygenic risk scores can be derived from summary-level data in the training and validation cohorts to estimate the heritability explained by SNPs and the number of causal variants84; a limitation of this approach is that SNPs are assumed to be uncorrelated, which can be approximately achieved by LD-pruning but precludes analyses of dense marker panels. The heritability explained by SNPs can alternatively be estimated from
