702
Genomic Prediction Within Family Combining Linkage Disequilibrium and Cosegregation Information
Dense genotypes at single-nucleotide polymorphisms (SNPs) have been widely applied in animal and plant breeding for predicting genetic merit for selection. Genomic best linear unbiased prediction and Bayesian linear regression models implicitly utilize cosegregation in addition to linkage disequilibrium (LD) between SNPs and causal variants. However, implicitly modeling cosegregation may not be sufficient to capture cosegregation information when the training set consists of multiple families or generations. Thus, a QTL model was proposed to explicitly account for both cosegregation and LD and evaluated with multiple families in the presence of either cosegregation, LD, or both. Results suggested that when cosegregation is the main source of genetic information in a quantitative trait, explicitly modeling cosegregation could result in substantially higher accuracy of genomic prediction within family, especially training with many families.
Keywords:
Genomic prediction
Linkage disequilibrium
Cosegregation
