A hybrid expectation maximisation and MCMC sampling algorithm to implement Bayesian mixture model based genomic...

Area of Science:

• Genomics
• Statistical Genetics
• Bioinformatics

Background:

• Bayesian mixture models are effective for genomic prediction and QTL mapping.
• Traditional Monte Carlo Markov Chain (MCMC) methods are computationally intensive for large genomic datasets.
• A novel hybrid approach, HyB_BR, combines Expectation-Maximisation with limited MCMC.

Purpose of the Study:

• To develop and evaluate an efficient computational method for genomic prediction and QTL mapping.
• To assess the HyB_BR method's performance against full MCMC in terms of accuracy and speed.
• To enable feasible genetic architecture inference in large-scale genomic studies.

Main Methods:

• Developed HyB_BR, a hybrid algorithm using Expectation-Maximisation and MCMC.
• Applied HyB_BR to dairy cattle quantitative traits and human disease data.
• Validated genomic prediction accuracy and compared computational time with full MCMC.

Main Results:

• HyB_BR achieved prediction accuracies comparable to full MCMC methods.
• Computational time was reduced by up to 17-fold (e.g., 45 hours to 3 hours).
• Identified similar significant SNPs and potential genes for traits/diseases as full MCMC.

Conclusions:

• HyB_BR provides equivalent prediction accuracy to full MCMC Bayesian models.
• The HyB_BR method significantly accelerates genomic prediction and QTL mapping.
• This efficient algorithm facilitates genomic prediction, QTL mapping, and genetic architecture inference in large datasets.