Using Interbull validation test results to improve national genomic evaluations

Abstract

The Interbull GEBV test software was recently updated, with new methods to account for genomic pre-selection of young AI bulls, and to add diagnostic features that help national evaluation units isolate and resolve issues if sub-optimal validation test results are observed. Applications of the updated Interbull tests to recent Canadian data confirmed genomic evaluations published by Lactanet are accurate and unbiased, particularly for traits with highest emphasis in current selection programs. Average GEBV validation slopes for Holstein were 1.06 across the three production traits, 1.06 for twenty conformation traits validated for internal use in Canada, and 1.07 for the five production and conformation traits requiring validation by Interbull. Validation slopes were notably lower for other traits, and especially for longevity, fertility, and calving traits when the test was based on a mixed population of foreign and local bulls and while ignoring expected bias in the early progeny-based evaluations of validation test bulls with a limited amount of progeny data. New features in the Interbull validation software were used to explore modeling improvements for traits with sub-optimal validation test results, and with direct longevity as an example trait to demonstrate proper uses of new diagnostic features now available in the updated Interbull validation software. Relative to initial validation test slopes for the mixed population of local plus foreign bulls, which averaged 0.73, 0.71, and 0.77 for the direct longevity, fertility, and calving traits evaluated by Interbull in GMACE, validation slopes for these same three groups of MACE traits were 0.98, 0.99 and 0.89 respectively, for local validation bulls born in Canada, when using an unbiased GEBV validation target. Low validation test slopes for GEBV were almost exclusively attributable to biases in the parent average contributions in GEBV, rather than genomic modeling of the young bull genotype effects, secondarily to the inclusion of foreign bulls born and pre-selected outside of Canada and then progeny-tested within Canada, and thirdly due to genomic pre-selection bias that is expected in a validation bull’s first GEBV that are based on only one year of progeny data. This research is ongoing, with plans to further explore validation of foreign bulls in Single Step evaluation systems that will include an integration of MACE information.