Interpreting SNP heritability in admixed populations

Abstract

SNP heritability (h2snp) is defined as the proportion of phenotypic variance explained by genotyped SNPs and is believed to be a lower bound of heritability (h2), being equal to it if all causal variants are known. Despite the simple intuition behind h2snp, its interpretation and equivalence to h2 is unclear, particularly in the presence of population structure and assortative mating. It is well known that population structure can lead to inflation in h2snp estimates. Here we use analytical theory and simulations to demonstrate that estimates of h2snp are not guaranteed to be equal to h2 in admixed populations, even in the absence of confounding and even if the causal variants are known. We interpret this discrepancy arising not because the estimate is biased, but because the estimand itself as defined under the random effects model may not be equal to h2. The model assumes that SNP effects are uncorrelated which may not be true, even for unlinked loci in admixed and structured populations, leading to over- or under-estimates of h2snp relative to h2. For the same reason, local ancestry heritability (hgamma) may also not be equal to the variance explained by local ancestry in admixed populations. We describe the quantitative behavior of h2snp and h2gamma as a function of admixture history and the genetic architecture of the trait and discuss its implications for genome-wide association and polygenic prediction.

Heritability within groups is uninformative about differences among groups: cases from behavioral, evolutionary, and statistical genetics

“…Despite the promise of the approach, heritability estimates based on local ancestry face their own challenges. Specifically, directional or stabilizing selection in the history of the source populations can lead to upward or downward biases in heritability estimates, respectively (Huang et al, 2023). In the supplementary text (section S2), we consider this problem and describe a related one that, in a pathological worst case, can lead to heritability estimates near 0 when the true heritability is 1.…”