StageWise

StageWise performs two-stage genomic selection analyses on multi-environment, multi-trait datasets to improve accuracy of genomic prediction and to optimize selection indices for plant breeding.

Key Features:

• Two-Stage Analysis: Leverages a two-stage analytical framework that uses the full variance-covariance matrix of genotype means from Stage 1 in Stage 2.
• Directional Dominance: Incorporates directional dominance models to account for heterosis and to model effects relevant to polyploidy and inbreeding depression.
• Multi-Trait Genomic Selection: Supports multi-trait genomic selection and the construction of selection indices, including restricted indices that constrain one or more trait responses to zero.
• Handling Polyploidy and Inbreeding: Models polyploid organisms and explicitly accounts for inbreeding depression in predictive analyses.
• Optimization of Selection Indices: Optimizes index coefficients by considering genetic correlations between traits to avoid undesirable correlated responses.
• Advanced Mixed-Model Techniques: Employs advanced mixed-model techniques for estimation and prediction in complex multi-environment, multi-trait contexts.
• Propagation of Stage 1 Error: Includes Stage 1 error structure in Stage 2 analyses via the Stage 1 variance-covariance matrix.
• Model Fit Evaluation (AIC): Uses Akaike Information Criterion (AIC) to assess model fit.

Scientific Applications:

• Plant Breeding: Applied to multi-environment trials to evaluate genotype performance across environments and traits for breeding decision support.
• Genomic Prediction: Enhances genomic prediction accuracy by modeling dominance, heterosis, polyploidy, and inbreeding effects in multi-trait contexts.
• Empirical Validation: Demonstrated on a potato dataset of 943 genotypes evaluated over six years, where inclusion of Stage 1 errors in Stage 2 reduced AIC for traits such as maturity, yield, and fry color.

Methodology:

Two-stage analysis using the full Stage 1 variance-covariance matrix of genotype means; directional dominance models; multi-trait genomic selection with (restricted) selection indices; modeling of polyploidy and inbreeding depression; optimization of indices using genetic correlations; advanced mixed-model estimation and prediction; model fit assessed with AIC.