Bayesian estimates of linkage disequilibrium

Area of Science:

• Genetics
• Bioinformatics
• Statistical genomics

Background:

• The maximum likelihood estimator for D', a linkage disequilibrium measure, exhibits bias towards disequilibrium, particularly in small sample sizes and with rare haplotypes.
• This bias can lead to inaccurate assessments of genetic associations.

Purpose of the Study:

• To develop a Bayesian estimation method for D' that corrects the bias inherent in the maximum likelihood estimator.
• To provide a more objective measure of linkage disequilibrium (LD) in the human genome.

Main Methods:

• Proposed a Bayesian estimation of D' using a prior distribution on pairwise SNP associations.
• The prior increases the likelihood of equilibrium with greater physical distance between SNPs.
• Utilized Markov Chain Monte Carlo (MCMC) methods for stochastic estimation and developed a numerical approximation for large datasets.

Main Results:

• The Bayesian estimator effectively corrects the bias towards disequilibrium observed in the maximum likelihood estimator.
• Demonstrated the feasibility of computing Bayesian estimates using MCMC and numerical approximations.
• The proposed method allows for more accurate estimation of LD patterns in large SNP datasets.

Conclusions:

• The Bayesian D' estimator offers a more objective assessment of linkage disequilibrium across the human genome.
• This improved accuracy leads to a more realistic identification of tagging single nucleotide polymorphisms (SNPs).
• Facilitates more powerful and precise genome-wide association studies (GWAS).