Current approaches to heritability estimation proceed by phenotyping many closely related individuals with a known genetic relationship, such as monozygotic (MZ) and dizygotic (DZ) twins7. Yang et al.10 avoided the use of related individuals by applying linear mixed models to estimate the heritability explained by genotyped SNPs (hg2). hg2 corresponds to the fraction of phenotypic variation that could be captured by hgwas2 under an additive model if GWAS sample sizes were infinitely large. While current estimates of hg2 are often much larger than hgwas2, they are typically only slightly more than half of h2 11. One reason hg2 is less than h2 is because hg2 does not include the contribution of variants poorly tagged by the genotyping platform, such as rare variants. Another reason for the difference in heritability estimates is that existing methods for estimating h2 can be biased12,13, since they rely on related individuals. As a result, epistatic interactions between SNPs, gene environment interactions, and the shared environmental factors of related individuals can all lead to inflated estimates of h2 12,13. We recently showed that by jointly using
