Updated Interbull software for genomic validation tests

Abstract

The Interbull GEBV test software was updated by adding new options and modernized methods for the validation of genomic predictions being used currently to select and market dairy bulls internationally.  Mis-aligned scales of evaluation can now be detected by the program, and the scales of full versus reduced data properly aligned before conducting validation tests, to reduce bias in the validation test results.  Genomic pre-selection bias can also be reduced by using new validation target options added to the program.  The new methods were applied to Canadian genomic evaluations with full data from August 2022, and reduced data equivalent to August 2018, for 30 MACE traits evaluated in Canada for each of the Holstein (HOL), Jersey (JER), and Ayrshire (RDC) breeds.  Evaluation scale alignments were needed for these data and reduced the bias of validation test slopes by as much as 10% for some traits.  Validation slopes were generally closer to 1 for all traits and breeds when using genomic validation targets that do not include GPS bias.  However, the smaller deviations from desired regressions close to 1.0 were partly due to higher auto-correlations with GEBV in reduced data, when using a genomic validation target instead of the current target that does not include genomic information.  New output files can be used to isolate data or modeling issues that might be causing poor validation test results, and thereby facilitate modeling improvements.  A new suite of validation tests was made possible by the software updates, which could be based on tests for bias in the average evaluations of animal groups.