Selection response in Tribolium to an index modified by introducing between‐family or within‐family or both restrictions in two steps

Abstract

IntroductionA fundamental strategy in selection programs is to combine maximum rate of response and minimum rate of inbreeding, these goals being in conflict with each other. Maximum selection response can be achieved at a cost of erosion in the effective number of breeding animals (a measure of the inbreeding level); reciprocally, the maximum effective number under selection can be preserved with a low response. The simultaneous consideration of both factors makes it difficult to decide on the use of individual (more effective in conserving effective number) or combined selection (maximizes response but yields low effective size).Q uinton et al. (1992) showed that comparing selection methods at the same level of inbreeding, rather than at the same selection intensity, changes the perspectives of current selection theory. If low to moderate inbreeding levels are considered, then phenotypic selection can yield higher response than selection on more accurate methods. Different methods have been proposed for maximizing selection response at the same level of inbreeding, i.e. to restrict the number of close relatives selected (N icholas and S mith 1983), to use false high heritability estimates in the genetic evaluation (G rundy and H ill 1993), to use assortative (S mith and H ammond 1986) or compensatory (G rundy et al. 1994) matings, to adjust estimated breeding values for the relationship with the already selected ones (G oddard and S mith 1990), to avoid matings of related individuals (T oro and P erez‐E nciso 1990), or to use factorial rather than hierarchical matings (W oolliams 1989; L eitch et al. 1994). Q uinton and S mith (1995) compared the merits of these methods using stochastic simulation; they concluded that none of the methods was best over all conditions, and that the use of false high heritabilities, or adjusted estimated breeding values with the relationships, does not seem to be recommended; besides, mating together those individuals with the lowest relationship has little effect on the accumulated inbreeding. W ray and G oddard (1994), and B risbane and G ibson (1995) indicated that if Gn is the genetic mean after n generations of selection and Fn is the mean inbreeding coefficient, a reasonable selection objective is Gn − DFn, where D is the value of a unit of inbreeding relative to a unit of genetic gain. M euwissen (1997) showed that these methods do not guarantee maximum genetic gains at some level of inbreeding and presented a rule for maximizing the genetic response with a predefined rate of inbreeding. His algorithm can be used to put a constraint on the variance of the selection response by replacing the additive relationship matrix by the prediction error variance (W oolliams and M euwissen 1993).W ei (1995a) developed a restricted phenotypic selection by considering limits on the number of individuals that will be selected from a family or on the family number selected. This less sophisticated method balances response and inbreeding. A restriction on the family number may lead to an increased response (but a decreased effective size), whereas restricting the proportion of selected individuals from a family is an efficient way to control the inbreeding (decreased response). W ei (1995b) generalized the method by introducing both restrictions. In this study, rates of response were compared under between‐family, within‐family, or both restrictions for a two‐trait selection index in a short‐term experiment with Tribolium.