Accurate estimation of heritability in genome wide studies using random effects models

Area of Science:

• Genetics
• Biostatistics

Background:

• Random effects models offer a novel approach for Genome-Wide Association Studies (GWASs).
• Previous methods, like Yang et al. (2010), estimated trait heritability but relied on a heuristic for model estimation.
• This heuristic approach may limit the accuracy and stability of heritability estimates.

Purpose of the Study:

• To develop a more stable and accurate maximum-likelihood (ML) estimation method for random effects models in GWASs.
• To improve upon the estimation techniques used by Yang et al. (2010).
• To determine the proportion of causal markers and understand their contribution to heritability.

Main Methods:

• The study adopts the random effects model framework from Yang et al. (2010).
• A novel maximum-likelihood (ML) estimation method is developed using Monte-Carlo Expectation-Maximization (MCEM).
• The MCEM approach incorporates a Markov Chain Monte Carlo (MCMC) method in the E-step for enhanced estimation.

Main Results:

• The developed MCEM-based ML method provides more stable and accurate heritability estimations compared to the heuristic approach.
• The method successfully estimates the proportion of causal genetic markers.
• This allows for differentiation between heritability arising from a few strong genetic factors versus many weaker ones.

Conclusions:

• The MCEM-based ML method offers a significant advancement in heritability estimation for GWASs.
• This refined approach enhances the understanding of genetic architectures underlying complex traits.
• The method provides a more robust tool for genetic research and heritability analysis.