Three statistical approaches for genetic analysis of disease resistance to vibriosis in Atlantic cod (Gadus morhua L.)1

Abstract

The objective of this study was to estimate and compare variance components and sire breeding values for disease resistance to vibriosis in Atlantic cod (Gadus morhua L.) using 3 statistical approaches. A total of 3,576 individually tagged juvenile cod from 50 full-sib families were infected with Vibrio anguillarum, which causes vibriosis, a frequently reported disease in cod aquaculture. The experimental fish were progeny of captured wild cod from populations of southern coastal cod (POP1), and northern coastal cod and northeast Arctic cod (combined as POP2 in the genetic analyses). Fish were randomly assigned to 1 of 3 test tanks, and daily mortality was recorded until the termination of the experiment at d 31 postinfection. Variance components were estimated separately for the 2 populations using a Cox regression model, univariate linear model, and a linear model that accounted for censoring. With all approaches, the additive genetic sire variance estimated from POP1 was greater than for POP2. Heritability estimates across populations varied from 0.08 to 0.17 depending on the method used. The Cox regression model and univariate linear model resulted in greater heritability estimates for POP1 (0.10 and 0.16) than for POP2 (0.08 and 0.13), whereas the contrary was true with a linear model that accounted for censoring (0.17 vs. 0.14). The predicted breeding values for the sires from the 3 approaches were highly correlated (0.97 to 0.99). This is likely due to the fact that censoring only occurred at the end of the test; i.e., observations of the most resistant fish were censored. The considerable genetic variation found in this study suggests that vibriosis resistance may be improved through selective breeding. The univariate linear model, even without censoring of the data, was robust for the estimation of breeding values using the present data. Therefore, inclusion of vibriosis resistance in the multivariate linear estimation of breeding values for the traits of economic importance in Atlantic cod seems appropriate.