Canonical-correlation analysis applied to selection-index methodology in quails

anglais. peer reviewed

anglais. Genetic evaluations in dual-purpose quails (Coturnix coturnix) have demonstrated that overall genetic gains in a breeding program are achieved not only based on a specific trait, but on several. The most common technique to use all this information is the selection index. Another alternative may be the canonical-correlation analysis applied to selection index. There is, however, a lack of studies using canonical correlation in quails. Hence, the objectives of this study were to apply canonical-correlation analysis to estimate the relationship of nine traits and to compare genetic gains obtained by this methodology to desired-gain selection index in three lines of quails. Data for three lines of layer quails consisted of body weight at 28 days (W28), egg weight (EW), age at first egg (AFE) and egg production at 30, 60, 90, 120, 150 and 180 days after onset of lay. Two sets of traits were established: the first one contained predictor variables (W28, EW and AFE) and the second one contained variables related to egg production. A selection index was constructed using the standardized coefficients of canonical covariates as weighting factors when a given canonical correlation was significant. We constructed two desired-gain selection indices: DG-SI1 and DG-SI2. The difference between them is that DG-SI2 had a desired gain for body weight set to 0. The estimated canonical correlations were as follows: 0.811, 0.058 and 0.003 for the yellow, 0.821, 0.181 and 0.076 for the red, and 0.825, 0.117 and 0.038 for the blue line. Only the first pair of canonical variates was significant (P<0.05). AFE and early stages of egg production were very influent and showed great importance in defining the canonical variates and, consequently, the estimated canonical correlations. All lines had, in general, similar results for the canonical analysis indicating that traits that drive management decisions in these lines would be the same. The indices under study showed differences in response to selection; however, they generally resulted in consistent favorable genetic gains. For all lines, the canonical selection index resulted in the lowest AFE and highest egg production at 30 days. The DG-SI1 showed the highest genetic gains for W28 in all lines. There was a general lower genetic gain of other traits for DG-SI1 at the expense of the desired genetic gain for W28. Selection for AFE, according to the canonical-correlation analysis, would have a great impact on the number of eggs produced. Canonical selection index is a good alternative for a desired-gain selection index. © 2014 Elsevier B.V.