Bayz

Bayz estimates region-specific genomic variances, covariances, and correlations using Bayesian multi-trait models to dissect complex trait architecture across diverse populations.

Key Features:

• Data import and model specification: Imports and describes data files and allows specification of analytical models based on response variables, link functions, and factors or covariates.
• Bayesian Statistical Framework: Employs Bayesian methods for statistical inference on genomic data.
• Multi-Trait Random Regression Model: Utilizes a multi-trait random regression model that incorporates latent variables to model heterogeneous variance and covariance across the genome.
• Region-Specific Analysis: Estimates genomic variances, covariances, and correlations at multiple genomic levels from whole genomes to specific chromosomes or SNP clusters (e.g., every 100 SNPs).
• Cross-Population Analysis: Handles data from different populations such as Chinese and Nordic Holstein cattle to analyze genetic correlations between traits across groups.
• Comparative Performance: Produces results comparable to multi-trait GBLUP when the entire genome is considered a single region.
• Identification of Key Genomic Regions: Identifies chromosomes or regions that contribute significantly to genomic variance and covariance for traits such as milk yield, fat yield, and protein yield.
• Correlation Analysis: Assesses positive and negative genomic correlations across populations, traits, and regions.

Scientific Applications:

• Quantitative genetics and genomics: Dissects complex trait architectures by estimating region-specific genomic parameters.
• Genomic prediction improvement: Supports improving genomic prediction accuracy through joint reference datasets from diverse populations.
• Livestock cross-population studies: Enables analysis of genetic correlations and region-specific effects in populations such as Chinese and Nordic Holstein cattle for traits like milk, fat, and protein yields.

Methodology:

Performs Bayesian inference using a multi-trait random regression model with latent variables to estimate variances, covariances, and correlations across genome regions while allowing model specification via response variables, link functions, and factors/covariates.