MegaBayesC

MegaBayesC applies Bayesian multivariate regression to improve genome-wide prediction accuracy and association study power by leveraging high-dimensional phenotype data.

Key Features:

• High-Dimensional Data Handling: Tailored for genomic analyses involving highly correlated and complex trait datasets across thousands of traits.
• Bayesian Multivariate Regression: Implements the BayesC prior within the R package MegaLMM to analyze multiple genetic variants and traits simultaneously.
• Integration with Hyperspectral Data: Demonstrated integration of hyperspectral reflectance data from 620 wavelengths to enhance genetic value prediction for grain yield in wheat.
• Simulation-Based Evaluation: Simulation studies assess estimation of quantitative trait loci (QTL) effect sizes across diverse genetic architectures and trait correlations.
• Two-Stage Whole-Genome Marker Handling: Employs a two-stage approach to manage whole-genome marker data efficiently.

Scientific Applications:

• Genomic Prediction: Incorporates diverse phenotype data to improve accuracy of predicting genetic values.
• Genome-Wide Association Studies (GWAS): Estimates effect sizes of QTL across various genetic architectures and trait correlations in simulation studies.
• Wheat Grain Yield Prediction: Uses hyperspectral reflectance (620 wavelengths) to enhance prediction of genetic values for grain yield in wheat.
• Flowering-Time Association in Arabidopsis thaliana: Applied to expression data from 20,843 genes to identify 15 SNPs associated with flowering time, 13 of which were within 100 kb of known flowering-time related genes and showed higher validation compared to single-stage analyses.

Methodology:

MegaBayesC uses a two-stage procedure consisting of preliminary candidate marker selection followed by multivariate regression using the BayesC prior implemented in the MegaLMM R package.