This study highlights biologically meaningless negative contributions to heritability estimates and introduces SAFE-h2 method and application tool for accurate genetic estimates in genome-wide association studies. By examining 74 phenotypes across eight species, SAFE-h2 revealed an average negative contribution, i.e., downward bias, of 30 units to SNP heritability estimates. Furthermore, SAFE-h2 improved the SNP heritability estimates by up to 56 units through capturing intra-locus (Ref-alleleSNPi × Alt-alleleSNPi) additive and non-additive effects within the framework of linear models across 50 phenotypes in six species. SNPs with potentially negative contributions to the heritability estimate should be excluded from the model. The widespread assumption that SNP-specific non-additive allelic effects are negligible does not rest on any inherent genetic or biological principles. The advantage of SAFE-h2 allelic adjustment approach is its reliance on a linear association model that jointly capture additive, dominance, overdominance, and heterosis-like effects. SAFE-h2 not only minimizes false-negative rates but also provides a safe p value interval to minimize false-positive SNP hits.
Darbani et al. (Thu,) studied this question.