Symposium review: How to implement genomic selection

摘要

Genomic selection was adopted very quickly in the10 yr after first implementation, and breeders continueto find new uses for genomic testing. Breeding valueswith higher reliability earlier in life are estimated bycombining DNA genotypes for many thousands of lociusing existing identification, pedigree, and phenotypedatabases for millions of animals. Quality control forboth new and previous data is greatly improved bycomparing genomic and pedigree relationships to correctparent–progeny conflicts and discover many additionalancestors. Many quantitative trait loci and genetests have been added to previous assays that used onlyevenly spaced, highly polymorphic markers. Imputationnow combines genotypes from many assays of differingmarker densities. Prediction models have graduallyadvanced from normal or Bayesian distributions withintrait and breed to single-step, multitrait, or other morecomplex models, such as multibreed models that maybe needed for crossbred prediction. Genomic selectionwas initially applied to males to predict progeny performancebut is now widely applied to females or evenembryos to predict their own later performance. Theinitial focus on additive merit has expanded to includemating programs, genomic inbreeding, and recessive alleles.Many producers now use DNA testing to decidewhich heifers should be inseminated with elite dairy,beef, or sex-sorted semen, which should be embryo donorsor recipients, or which should be sold or kept forbreeding. Because some of these decisions are expensiveto delay, predictions are now provided weekly insteadof every few months. Predictions from internationalgenomic databases are often more accurate and costeffectivethan those from within-country databases thatwere previously designed for progeny testing unless localbreeds, conditions, or traits differ greatly from thelarger database. Selection indexes include many newtraits, often with lower heritability or requiring largeinitial investments to obtain phenotypes, which providefurther incentive to cooperate internationally. The genomicprediction methods developed for dairy cattleare now applied widely to many animal, human, andplant populations and could be applied to many more.