4.5. Alternative breeding strategies
In this study, males used to produce animals to the grow-out tier were selected from the nucleus based on TMI, while the females were selected from the multiplier tier, based on CAND phenotypes. Other strategies of selection could also be possible. One strategy could be to use TMI selection of females at the multiplier level. This would increase the selection accuracy of the dams, and therefore the accuracy of the selection in the multiplier tier. Another strategy that we tested (results not shown) was schemes with the same selection proportions (10 males and 600 females used in the grow-out tier), but with selection based on phenotypic selection of both sexes and compared it to TMI selection of both sexes with recording of the sib-trait on the females also in the multiplier tier. The results were as expected: TMI selection of both sexes gave the highest profit, on average it was 8.61 σBV. Phenotypic selection gave on average a profit of 5.95 σBV. The optimal scheme recommended above (Trunc_intense) gave in comparison an average profit of 6.93 σBV. A strategy based on TMI for both sexes was more advantageous when rg was zero or negative. When the correlation was positive the strategies performed more similar. However, it is questionable whether this method can be recommended since having an accurate estimate of TMI with a sib-test in the multiplier tier would be very costly and it requires a change in the organisation of the multiplier tier. A third strategy could be to use overlapping generations on the dams if the species is a multiple spawner. For other species with for instance growth measured on an early stage or where it is possible to do a disease test on fingerlings, there would be information on how the offspring perform early, and therefore possible to save the best dams and reuse them. It will then be possible to give a better estimate on how the offspring will perform, since there is available information from how older full and half sibs have performed. Then the accuracy on the dams will increase, and thus it could be possible to take higher risk by the use of fewer dams. Genomic selection (Meuwissen et al., 2001) could also be used within each of the tested strategies to increase the accuracy especially for traits that are not measured on the selection candidates, and thereby for instance give females from the multiplier tier breeding values for both traits. This would, however, require genotyping of a large number of fish, both selection candidates and fish with phenotypes, and would hence be a rather expensive strategy, most suited for traits of high economic importance that are expensive to measure phenotypically.