Foundation Animal Effects

UNDER CONSTRUCTION...

Implementing BLUP animal models without accounting for differences in groups of foundation animals' genetic merit can result in biased EPDs. Differences in additive genetic merit between groups of foundation animals can be a result of different breeds of origin, founders of the same breed entering the data used for genetic evaluation at different periods of time or a combination of both. The differences from entering the data at different times are due to genetic trend resulting from within breed selection.

Additive genetic groups

When additive genetic group effects are not included all foundation animals are modeled as coming from a population with the same average genetic merit . Including additive genetic groups permits the model to predict genetic differences that occur between the groups. These differences impact the EPDs of the descendants of the foundation animals.

When designating genetic groups by periods of time there is a balance between size of group and amount of time on which to cluster groups. Too short a period results in groups predictions with high prediction errors. Too long a time period decreases the precision of the prediction. This problem is compounded in multi-breed evaluations where groups are designated with both breed of founder and generation.

A founder animal can be assigned to more than one genetic group in a multi-breed evaluation. The genetic group assignment is determined by the animal's parents' breed group of origin. Additionally, if one parent is known (often the dam) then the animal is partially assigned to the breed of sire or year genetic group.

Additive genetic groups can be included as fixed effects by absorbing the groups equations into the inverse numerator relationship matrix. Direct simplified computational procedures have been well described. While computationally easy to set up, including additive genetic groups may cause instability during iteration for the solution of the mixed model equations. More sophisticated preconditioners can be useful to improve performance to convergence during iteration. When implemented this way the resulting breeding value predictions include the fixed group effects.

Alternatively, additive genetic groups can be included as covariates. Doing so will improve iterative performance. The covariates will be a set of very dense equations (many non-zero values). However, modern computers with long vector processing capability can process these large and dense problems effortlessly. The resulting breeding value predictions will not include the fixed group effects. These effects must be added back to the predictions so that they include all the additive genetic variance in the model.

Fitting groups as fixed effects results in a non-singular set of equations. An appropriate constraint to the linear system of mixed-model equations must be applied.

Metafounders

The use of genomic data allows for the inclusion of an alternative to additive genetic groups. Unlike additive genetic groups, metafounder effects can account for relationships between foundation groups even though they may be distinct breeds. Because all breeds ultimately come from the same origins they have many of the same alleles.

Genomic data can be used to estimate the alleles common by descent from an original ancestral population. The metafounder approach uses a "pseudo-individual" to represent an animal of ancestral origin analogous to additive genetic groups. These pseudo individuals' relationships are estimated using pedigree and genomic data across populations. The pseudo individual then contributes to the relationship matrix calculated from the pedigree data used in the mixed model equations.