CodataGS

CodataGS integrates external biological information about genetic markers into genomic selection models to improve prediction accuracy for quantitative traits using a hierarchical generalized linear model framework.

Key Features:

• Integration of External Marker Information: Incorporates external biological data about markers, including location and uncertainty of quantitative trait loci (QTLs), into genomic selection models.
• Hierarchical Generalized Linear Model (hglm) Framework: Operates within an hglm framework to accommodate varying degrees of marker information uncertainty and genetic architectures that deviate from infinitesimal models.
• Heteroscedastic Random Effects and Link Function: Models marker-specific variances using a link-function approach and allows heteroscedastic random effects across markers.
• Efficiency and Scalability: Optimized for scenarios where the number of markers far exceeds the number of individuals and provides a speed advantage over existing hglm implementations.
• Improved Prediction Accuracy: Demonstrates simulated-data accuracy improvements of 3.8%–23.2% over SNP-BLUP depending on trait genetic architecture and precision of external information.

Scientific Applications:

• Animal breeding: Enhances genomic selection and trait prediction in livestock breeding programs by incorporating marker-level biological information.
• Plant genetics: Refines genomic prediction and selection in crop and plant genetics using external QTL and marker data.
• Genomic selection research: Supports studies comparing modeling approaches and assessing the impact of external marker information on prediction accuracy.

Methodology:

Fits linear mixed models within a hierarchical generalized linear model (hglm) framework, using heteroscedastic random effects whose variances are modeled by a linear predictor with a log link to incorporate external marker information via a link-function approach.