Dissect high-resolution genetic architecture of complex phenotypes by accurately estimating gene-based conditional heritability

Abstract Despite many methodological studies on the global heritability estimation of complex traits, very few methods have been designed to dissect local heritability accurately. A precise local heritability would provide a high-resolution map for precision medicine. We report an effective heritability estimator (EHE) for local heritability estimation using p values from genome-wide association studies (GWAS). With a directly converting marginal heritability to the effective one, EHE shows higher accuracy and is the only method correctly calculating standard errors of local heritability estimation among the four compared methods. Importantly, EHE can be directly used to estimate the conditional heritability of nearby regions or genes in which the redundant heritability can be cleaned. The conditional heritability estimation procedure can be supervised by tissue-specific expression profiles to prioritize and quantify functionally more important genes of complex phenotypes. Applying EHE to 42 complex traits from the UK biobank, we found that heritability density in lncRNA genes tends to be higher than protein-coding genes, and genes pleiotropic to more phenotypes are not enriched for more heritability among candidate susceptibility genes. EHE provides an efficient way to finely and precisely dissect the genetic architecture of complex phenotypes.