Realized Genome Sharing in Heritability Estimation Using Random Effects Models

Area of Science:

• Quantitative genetics
• Statistical genetics

Background:

• Heritability estimation is crucial for understanding genetic contributions to traits.
• Random effects models are commonly used, but their accuracy depends on correct model specification.
• The "missing heritability" problem highlights discrepancies between estimated and actual genetic influence.

Purpose of the Study:

• To provide formulas for the limiting distribution of the maximum likelihood estimate in a two-component random effects model for heritability.
• To assess the impact of using incorrect kinship measures on heritability estimation.
• To investigate how model mis-specification, particularly of the genetic correlation matrix, contributes to the missing heritability problem.

Main Methods:

• Derivation of formulas for the limiting distribution of the maximum likelihood estimate.
• Theoretical analysis of asymptotic sampling variance and bias.
• Simulation study using a population-based design and the genomic relationship matrix (GRM).

Main Results:

• Formulas are provided for heritability estimation, valid even with incorrect kinship measures.
• Asymptotic sampling variance depends on study design and variation in the kinship measure.
• Mis-specified correlation matrices lead to asymptotic bias, influenced by the difference between fitted and true matrices.
• Estimating heritability with population-based designs and the GRM may exacerbate the missing heritability issue.

Conclusions:

• The accuracy of heritability estimates is sensitive to the correct specification of kinship and genetic correlation matrices.
• Mis-specification, particularly using the GRM in population studies, can lead to biased heritability estimates and contribute to the missing heritability problem.
• These findings have implications for designing genetic studies and interpreting heritability estimates.