Effect of More Stringent Convergence Criterion of Estimated Breeding Values on Response to Selection

Abstract

Prediction of breeding values of individuals using an animal model BLUP evaluation takes substantial computing resources which can be a limiting factor in simulation studies. When a large set of equations has to be solved iteratively, computing time depends on the stringency of the convergence criterion (CC). This study investigated the interaction between computing time and selection response as convergence was varied for single and multiple trait (four traits) BLUP evaluation. Although the average true breeding values (TBV) of the top 2% of individuals improved with increasingly stringent CC, the gain in average TBV was not sufficient to justify the number of extra iterations needed at increasingly stringent CC with single trait BLUP and h²=0.25. Under multiple trait BLUP evaluation, the TBV of the selected group and hence genetic response was more affected by CC. The proportion of animals (PA) that were in the final selected group at the most stringent CC, was observed for all high values of the CC. PA was more dependent on CC, than the average TBVs, indicating that animals entering and leaving the selected group as iterations proceed are those around the truncation point. Under both single and multiple trait analysis, the BLUP subroutines took about 33 and 49% of total computing time respectively for 12110 individuals, when the most stringent CC was used.