SNP Allocation For Estimating Heritability (SAFE-h²): A tool to explore genomic origins of phenotypes for estimation of SNP heritability using additive-only allelic effects or additive and non-additive allelic effects

Abstract SNP heritability is a marked indication for genetic gains. Extreme polygenicity, i.e. , small effect-size causal SNPs, dictates as many SNPs as possible to be involved in heritability estimations. On the contrary, non-causal SNPs underestimate SNP heritability levels via negative contributions which is confirmed here on simulated and real datasets. Negative contributions to heritability are biologically absurd as polymorphic genomic positions can only have null or positive contributions to heritability. This paradox signifies a trade-off between the incorporation of unknown causal SNPs and the boycott of unknown non-causal SNPs for rigorous approximations of heritability. SAFE- h 2 (SNP Allocation For Estimating Heritability) is an application tool to overcome this dilemma by utilizing association p -values for SNP heritability profiling. SAFE- h 2 determines an association p -value threshold to filter out SNPs from its upper-bound as they contribute negatively to heritability estimations. Considering SNP heritability estimates which were calculated using all SNPs for 74 phenotypes in five plant and four animal species, SAFE- h 2 revealed up to 90 units (30 units on average) negative contribution imposed by SNPs from upper-bounds of p -value thresholds. SAFE- h 2 also provides a safe p -value interval to minimize false-positive SNP hits’ contributions to heritability estimates. Furthermore, SAFE- h 2 captures non-additive effects through intra-locus allelic adjustments. The allelic adjustment algorithm of SAFE- h 2 captured additional phenotypic variance and improved the SNP heritability estimations for random phenotypes assigned to the human genotypes as well as artificial phenotypes associated, at different rates, with a biallelic locus. Likewise, SAFE- h 2 revealed considerable uncaptured phenotypic variance when relying on additive effects and improved the SNP heritability estimations up to 56 units by joint analysis of intra-locus additive and non-additive allelic effects for 50 phenotypes in four animal and three plant species. We found the EMMAX algorithm to be the most stable and accurate model for SNP heritability profiling by SAFE- h 2 .