Uso de modelos lineares mistos na avaliação genética de escores visuais: estudo de simulação

• Autor(a) principal: Duitama, L.o.
• Data de Publicação: 2014
• Outros Autores: Farah, M.m., Utsunomiya, A.t.h., Ono, R.k., Pires, M.p., Fonseca, R.
• Tipo de documento: Artigo
• Idioma: por
• Título da fonte: Repositório Institucional da UNESP
• Texto Completo: http://dx.doi.org/10.1590/1678-6351; http://hdl.handle.net/11449/114177
• topic: gado de corte; parâmetros genéticos; software R; beef cattle; genetic parameters; software R

This study aimed to evaluate the application of linear mixed models methodology using simulated data for trait visual scores, two population structures (with and without selection), two levels of heritability (0.1 and 0.4) and four levels of connectability (8, 20, 38 and 60%) were examined. Populations with and without selection consisted of 6,660 and 3,360 animals respectively, of which, the last in 2460 had score visual phenotypes. The scores were simulated with an underlying normal distribution from which intervals were defined corresponding to each category of the visual scores. The simulation process was performed in software R, to estimate genetic parameters and predict breeding values through software Wombat using animal model, considering models with and without fixed effects. The evaluation criteria were: the mean squared error (EQM) for heritability and Spearman correlations between true and predicted breeding values. Estimates of heritability showed close to the true value in scenarios without selection (0.084-0.101 and 0.367-0.389), however when selection was applied heritability was underestimated (0.032 and 0.278). Consistent with heritability, the correlations were higher for the scenarios without selection and with heritability of 0.4 (0.86-0.89). In all scenarios simulated the inclusion of fixed effects in the model improved the estimates of heritability and correlations between true and predicted breeding values. The level of connectability affected the correction of fixed effects done by allocating visual scores. The linear mixed model methodology can be used in the estimation of genetic parameters and predicted breeding values for visual scores in populations without selection, however, is not suiting in populations under selection.