Optimization of dairy cattle breeding plans with increased female reproductive rates

IntroductionNicholas and Smith (1983) proposed Multiple Ovulation and Embryo Transfer (MOET) nucleus breeding schemes to increase reponse rates in dairy cattle breeding. Predicted genetic gains were up to twice as high as those of conventional progeny testing schemes. In the MOET nucleus breeding schemes, selection was within a closed nucleus herd using short generation intervals and mainly sib information. Juga and Maki-Tanila (1987) simulated MOET nucleus schemes and found that predicted rates of gain were 124 % higher than simulated. From this two questions arise: i) how to predict response rates correctly; and ii) how to make optimal use of MOET in dairy cattle breeding.Prediction of response ratesMajor factors decreasing response rates are:1. Reduction of variances due to selection. This consists of reduction of variances of information sources, which were previously selected for, and reduction of genetic variance due to linkage disequilibrium between genes as described by Bulmer (1971). The effect of linkage disequilibrium could be accounted for by correcting the genetic variance for all selection on ancestral sources of information. This factor reduced response rates by 20-30%.2. Reduction of selection differentials due to small numbers of selection candidates and small numbers of families (Hill, 1976). Selection differentials are often predicted assuming that breeding value estimates of selection candidates are uncorrelated. However, family relations between selection candidates cause correlations between breeding value estimates. Especially in schemes with short generation intervals, where breeding value estimates are mainly based on information of sibs, correlations between breeeding value estimates of sibs are high, since these are based on the same sources of information. An approximation for the reduced selection differentials in a nested full-half sib family structure was derived. Predicted response rates were reduced by up to another 30%.3. Variance reduction due to inbreeding. Also, this factor reduces genetic variance and thus genetic gain substantially, when the inbreeding coefficient is large. Even with an inbreeding rate of 5% per generation, i.e. effective population size is 10 animals per generation, it takes about 5 generations before the inbreeding coefficient is large enough to be of importance. Therefore, average selection response during the first five generations is not much reduced (6%). For 10 generations this figure is 13%. Thus, the impact of this factor depends on the time horizon (here: the time period during which the breeding population is expected to be closed to foreign breeding stocks). In view of the large difference in effective population size, i.e. 10 animals per generation vs. infinite (no inbreeding), it is concluded that up to 10 generations the impact of variance reduction due to inbreeding on the ranking of breeding schemes is not large. The first and second factor were accounted for in this study. Variance reduction due to inbreeding was neglected.Breeding schemesIn nucleus breeding schemes, nucleus dams are selected from the female nucleus population, which has the same genetic level as the bull stud. In progeny testing schemes, bull dams are selected from commercial herds. However, some cows born in commercial herds are daughters of bull sires and bull dams and are thus of equal genetic level as the bulls in the stud. These cows are comparable with the nucleus females in nucleus schemes and have an higher probability of being selected than 'normal' commercial cows. Thus, progeny testing schemes are open nucleus schemes, where daughters of bull sires and bull dams form the nucleus females and where 'normal' cows are the base population.It was assumed here, that milk production records were not biased by housing of nucleus animals, i.e. in special nucleus herds or dispersed across commercial herds. There are only three differences between the nucleus schemes proposed by Nicholas and Smith (1983) and progeny testing schemes, that use MOET to increase reproductive rates of bull dams:1. In the closed nucleus schemes of Nicholas and Smith, selection of dams is within the nucleus herd, whereas in open nucleus / progeny testing schemes nucleus and base population females serve as selection candidates. When, relative to the nucleus size, many bull dams have to be selected, this is advantageous for the open nucleus/progeny testing scheme. When the number of selected dams is small due to the use of MOET, relatively many dams will be selected from the genetically superior nucleus population. This implies that the open and closed nucleus schemes become more similar, when female reproductive rates increase. With on average 8 offspring per donor cow per year, differences in genetic gain between open and closed nucleus breeding schemes were small.2. Generation intervals are much longer in progeny testing schemes than in nucleus breeding schemes, which is partly due to progeny testing of bulls. James (1987) shows that generation intervals could be optimised in any schemes by selecting for BLUP breeding values across all age classes. The ad hoc nature of this optimization makes predefining generation intervals of breeding schemes redundant. Consequently, this difference between nucleus breeding and progeny testing schemes disappears. However, in practical progeny testing schemes, generation intervals are not optimised: only proven bulls (at least 5 years old) are considered for selection and usually cows without a milk production record are not considered for selection of bull dams. Selection response increases by about 15%, when these restrictions are abolished. Biasedness of breeding value estimates of young animals will reduce this improvement and selection response might be even reduced.3. Progeny testing of young bulls in the base population. In open nucleus breeding schemes with optimised generation intervals, progeny testing reduces genetic gains by up to 10% depending on the number of sires used.Variances of selection responsesVariance of the selection response is a measure for risk of the breeding plan. Further, variance of the selection response and inbreeding are positively related. Reduction of generation intervals due to the optimization procedure increased the standard deviation of the selection response by a factor 2 to 3. Utility theory was used to weigh selection response against its variance. A quadratic approximation of the utility function and maximum risk aversion were assumed. Schemes with optimised (short) generation intervals had the highest utility.Closed nucleus schemes had a lower utility than open nucleus schemes, both with optimized generation intervals. This was due to the 80 % higher standard deviation of the selection response in closed nucleus schemes. Differences in selection response were small: closed nucleus schemes had 3% more selection response than open nucleus schemes (8 offspring per donor cow). In these open nucleus schemes, selection of nucleus dams from the base was very intense, which resulted in less genetic variance in the nucleus offspring. This caused the small difference in genetic gain.Main conclusions- Variance reduction due to selection reduces predicted genetic gain by 20 - 30 %.- Correlations between breeding value estimates of relatives reduce predicted selection differentials in breeding schemes by up to another 30%.- As reproduction rates of females increase, optimised open nucleus schemes (or progeny testing schemes) become more closed. With an average of 8 offspring per selected cow, open and closed nucleus schemes have almost euqual genetic gains.- Variances of selection responses increase substantially, when generation intervals are reduced. However, when selection response and its variance were weighted, shorter generation intervals were still prefered.- Variance of the selection response of closed nucleus schemes is higher than that of open nucleus schemes (both having optimised generation intervals). Therefore, under the assumption that field and nucleus herd milk production records both are unbiassed, open nucleus schemes were prefered.