Accounting for residual correlations among regional genomic predictions via GMACE


  • P Sullivan


Despite encouraging preliminary results with GMACE, further research has identified some important problems. Errors can be expected in the arbitrary input parameters needed to approximate residual correlations, and GMACE results are quite sensitive to these errors. Restricting bulls to a single GEBV in S-GMACE addresses this concern by eliminating the need for within-sire residual correlations. However, fitting only within-sire residual correlations may be inadequate due to the accumulation of information among relatives via the relationship matrix. It is recommended to further restrict S-GMACE, by preventing the accumulation of genomic information among relatives, to expedite an international genomic evaluation service for young genomically-tested animals. Research should also continue to develop and test less-restricted and/or full-scale GMACE applications