A mating advice system in dairy cattle incorporating genomic information

摘要

This study investigated the effects of alternative matingprograms that incorporate genomic information onexpected progeny herd performance and inbreeding, aswell as methods to include un-genotyped animals in suchmating programs. A total of 54,535 Holstein-Friesiancattle with imputed high-density genotypes (547,650SNP after edits) were available. First, to quantify theaccuracy of imputing un-genotyped animals (often anissue in populations), a sub-population of 729 genotypedanimals had their genotypes masked, and theirallele dosages were imputed, using linear regression exploitinginformation on genotyped relatives. The referencepopulation for imputation included all genotypedanimals, excluding the 729 selected animals and theirsires, dams, and grandsires, and had either (1) theirsires’ genotypes, (2) their dams’ genotypes (3) boththeir sires’ and their dams’ genotypes, or (4) both theirsires’ and maternal grandsires’ genotypes introducedinto the reference population. The correlations betweentrue genotypes and the imputed allele dosages rangedfrom 0.58 (sire only) to 0.68 (both sire and dam). Aherd of 100 cows was then simulated (1,000 replicates)from the sub-population of 729 imputed animals. Thetop 10 bulls from the genotyped population, based ontheir total genetic merit index (TMI) were selected tobe used as sires. Three mating allotment methods wereinvestigated: (1) random mating, (2) sequential matingbased on maximizing only the expected TMI of theprogeny, and (3) linear programming to maximize agenerated index constructed to maximize genetic meritand minimize expected progeny inbreeding as well asintra- and inter-progeny variability in genetic merit.Relationships among candidate parents were calculatedusing either the pedigree relationship matrix or thegenomic relationship matrix; the latter was constructedusing either the true genotypes of both parents or thetrue genotypes of the sire plus the imputed allele dosagesof the dam. Using the genomic co-ancestry estimatesresulted in lower average herd expected genomicinbreeding levels compared with using the pedigreebasedco-ancestry estimates. Additionally, if the damswere not genotyped, using their imputed allele dosagesalso resulted in lower average herd expected inbreedinglevels compared with using the pedigree co-ancestryestimates. The inter-progeny coefficient of variationfor selected traits, milk and fertility, estimated breedingvalues were reduced by 12 to 65% using the linearprograming method compared with sequential mating.