Heritability for body weight at harvest size in the Pacific white shrimp, Penaeus (Litopenaeus) vannamei, from a multi-environment experiment using univariate and multivariate animal models

To estimate family BLUP breeding values and the heritability of body weight at harvest size (BW) in the Pacific white shrimp, Penaeus (Litopenaeus) vannamei, an experiment was conduced using information from two farm units of a Mexican hatchery and two shrimp population densities at each location. Data consisted of 12,658 shrimps that were siblings from 48 sires and 77 dams with a nested dam–sire structure. Shrimps were individually weighed at an average age of 130 days post-hatching. BW phenotypic mean (S.D.) was 18.2 (2.4) g, with values ranging from 8.4 to 30.0 g. Data were analyzed using univariate and multivariate models that considered BW within location by density pond environment as a different trait and included or not a common full-sib effect (c). The multivariate animal model included fixed effects of days from hatching and sex. For univariate models that included c effects, BW heritability (S.E.) estimates ranged from 0.24 (0.14) to 0.35 (0.18) across environments (heritability was zero in one environment). For multivariate models (excluding the environment with zero heritability) the heritabilities increased and ranged from 0.37 (0.06) to 0.45 (0.09). Standard errors of heritabilities and c effects were both drastically reduced in the multivariate analysis. Pairwise genetic correlations between environments were from 0.80 (0.08) to 0.86 (0.04). These differences may be indicative of genotype–environment interaction for BW at 130 days post-hatching. Statistical problems found to separate c from additive genetic effects both in univariate models were reduced using multivariate models. Correlation between family raw phenotypic means and family BV means from the multivariate analysis was 0.93 indicating a rather low risk of miss selecting superior families if BLUP solutions were neglected using replicated environment data. It is also concluded that use of incorrect statistical models or unreplicated data may lead to biased or inaccurate estimates of genetic parameters in shrimp breeding programs.