Efficient Estimation of Marker Effects in Plant Breeding

Area of Science:

• Agricultural Science
• Genetics
• Bioinformatics

Background:

• Genomic-enabled selection (GES) is vital for plant breeding.
• Evaluating prediction models is key for successful GES implementation.
• Computational time and predictive ability are critical metrics for genomic prediction pipelines.

Purpose of the Study:

• To introduce novel methods for genomic prediction that balance high accuracy with computational efficiency.
• To evaluate the predictive ability and marker effect estimation of proposed methods.
• To compare the proposed single-stage approach with existing genomic prediction techniques.

Main Methods:

• Developed a non-Markov Chain Monte Carlo (MCMC) method using a Laplace prior for marker effect estimation.
• Implemented an iterative framework for single-stage whole-genome regression with replicated observations.
• Conducted cross-validation studies on simulated plot-level data with unbalanced structures.
• Compared methods including genome-wide association analysis (GWAS), whole-genome regression (WGR), and GBLUP.

Main Results:

• The proposed methods demonstrated high computational efficiency and robust prediction across datasets.
• Accurate estimation of marker effects was achieved compared to GWAS and WGR methods.
• The single-stage approach showed competitive performance against GBLUP and two-stage methods, especially with unbalanced data.

Conclusions:

• The developed methods offer a promising framework for efficient and accurate genomic prediction in plant breeding.
• These advancements can accelerate the implementation of genomic selection for crop improvement.
• The single-stage approach provides a valuable alternative for handling complex breeding data structures.