A single-step genomic model with direct estimation of marker effects

Area of Science:

• Animal Breeding and Genetics
• Quantitative Genetics
• Genomic Prediction

Background:

• Multi-step genomic models are widely used but can be less accurate and susceptible to genomic preselection bias.
• Single-step genomic models offer improved accuracy by integrating all available data.
• Efficient computational methods are crucial for implementing advanced genomic evaluation models.

Purpose of the Study:

• To introduce a novel single-step single nucleotide polymorphism (SNP) model for genomic evaluation.
• To develop efficient algorithms for solving the equations of this new model.
• To enable direct estimation of SNP effects and flexible modeling of their contributions.

Main Methods:

• Proposed a single-step SNP model incorporating an additional step for estimating SNP marker effects.
• Developed efficient computing algorithms, including preconditioned conjugate gradients and a special updating algorithm.
• Extended the model to multiple-trait cases using block-diagonal (co)variance matrices for efficient multivariate SNP effect estimation.

Main Results:

• The single-step SNP model allows flexible modeling of SNP effects and includes a residual polygenic effect to reduce prediction inflation.
• Efficient algorithms were developed for estimating SNP effects and separating them from polygenic effects.
• A general prediction formula was derived for rapid, interim genomic evaluations of candidates without phenotypes.

Conclusions:

• The proposed single-step SNP model enhances genomic evaluation accuracy and computational efficiency.
• The model provides flexibility in estimating SNP effects and correcting for genomic preselection.
• Implementation in large populations like Holstein with across-country reference populations is feasible and discussed.