Measuring missing heritability: inferring the contribution of common variants

Area of Science:

• Genetics
• Biostatistics
• Epidemiology

Background:

• Genome-wide association studies (GWASs) identify genetic variants linked to diseases, but often explain little heritability.
• The "missing heritability" problem may stem from numerous small-effect common variants undetectable by current GWAS sample sizes.
• Accurate heritability estimation is crucial for understanding genetic contributions to disease.

Purpose of the Study:

• To evaluate the performance of restricted maximum likelihood (REML) for heritability estimation in disease studies.
• To develop and validate a novel statistical framework for unbiased heritability estimation from common variants.
• To quantify the proportion of heritability attributable to common variants across various diseases.

Main Methods:

• Demonstrated underestimation of heritability by REML in case-control disease studies, particularly with rare diseases and larger sample sizes.
• Developed phenotype correlation-genotype correlation (PCGC) regression, a generalized method for heritability estimation.
• Applied PCGC regression to six diseases to estimate the heritability contributed by common variants.

Main Results:

• REML significantly underestimates heritability in disease studies, with underestimation increasing with disease rarity, heritability, and sample size.
• PCGC regression provides unbiased heritability estimates.
• Common variants account for 25%-56% of phenotypic variance and 41%-68% of heritability (mean 60%) across six studied diseases.

Conclusions:

• Common variants likely explain a substantial portion, at least half, of the heritability for many common diseases.
• PCGC regression offers a robust and generalizable method for heritability estimation, applicable to various study designs and covariates.
• Findings necessitate re-evaluation of the role of common variants in disease etiology and heritability.