Logotipo do repositório
 

Publicação:
Application of ssGBLUP for variance components estimation and genomic prediction in two Zebu cattle populations

Carregando...
Imagem de Miniatura

Orientador

de Albuquerque, Lucia Galvão

Coorientador

Rey, Fernando Sebastian Baldi

Pós-graduação

Ciência Animal - FCAV

Curso de graduação

Título da Revista

ISSN da Revista

Título de Volume

Editor

Universidade Estadual Paulista (Unesp)

Tipo

Tese de doutorado

Direito de acesso

Acesso restrito

Resumo

Resumo (inglês)

The ssGBLUP combines genomic and pedigree information, which can increase the accuracy in the estimation of variance components (VCE). In cases such as the traits traditionally evaluated in commercial genetic improvement programs for Nelore cattle in Brazil, the use of ssGBLUP often involves mixing information from an old pedigree-based selection program with current genomic selection (GS) information. This is controversial because selection decisions are based on genomic information, and only subsets of animals are genotyped, which represents an additional selection step that can introduce bias into genomic estimated breeding values (GEBV). On the other hand, ssGBLUP can improve VCE in GS programs for breeds with fewer records, such as Brahman cattle, especially in traits with low heritability, such as gestational loss (PL), and when considering genotype-environment interaction (GxE). Therefore, the overall objective was to evaluate the application of ssGBLUP for VCE and genomic prediction of production and reproduction traits in two Zebu cattle populations. In the case of Nelore cattle, the first specific objective was to estimate variance components for growth and reproductive traits using two different relationship matrices (relationship matrix A based on pedigree and relationship matrix H based on pedigree and genomic information), as well as considering phenotypic records collected before or after the implementation of GS in Brazilian Nelore cattle. Additionally, we evaluated how GEBV are affected by VCE and by excluding records prior to the implementation of GS. In this case, the traits evaluated were weight at 120 days (W120), weight at 450 days (W450), and scrotal circumference at 450 days (SC450). Three datasets were used for VCE: 1) including all phenotypes (All), 2) only records of animals born before 2010 (Before), or 3) records of animals born since 2010 (After), according to the start of GS. We used matrix A in all three datasets (A_before, A_after, and A_all), while H was used in two situations: H_All and H_After. Different VCE were used for the prediction of GEBV through ssGBLUP. This stage used two possible datasets: all available records (Dataset 1) and only records collected since 2010 (Dataset 2). Validation was conducted using the LR method. For the three traits, heritability increased using A_After, and for W450 and SC450 heritability decreased with the use of matrix H. For W120, Dataset 1 and VCE from A_After showed the highest accuracy for direct and maternal GEBV. For W450 and SC450, Dataset 1 and VCE using H_After allowed high accuracy for GEBV. Furthermore, in both traits, using Dataset 2 did not add bias or dispersion and maintained high accuracy of GEBV. For W450 and SC450, genetic variance and heritability increased with GS. For W120, GEBV were more accurate using A for VCE. It was possible to obtain high accuracy of GEBV for W450 and SC450 using H in VCE and discarding records prior to GS. It is possible to exclude records prior to the implementation of GS without increasing bias and dispersion in predictions. In the case of Brahman cattle, the second specific objective was to estimate variance components and genetic parameters for PL and evaluate its genetic correlation with growth and reproductive traits assessed in a commercial genetic improvement program, comparing the use of matrices A and H. For this, PL records from three pregnancy orders (heifers, primiparous cows, and multiparous cows) were collected for two herds in Bolivia. Univariate and multivariate threshold models were fitted for PL in the three pregnancy orders, and multivariate models were also fitted between PL and traditional traits such as weights adjusted at 450 (W450) and 550 (W550) days of age, scrotal circumference adjusted at 450 (SC450) and 550 (SC550), cow accumulated productivity (ACP), age at first calving (AFC), and stayability (STAY). Matrices A and H were tested in all models. Heritability for PL was low in the three orders, but the use of matrix H increased heritability in heifers. High genetic correlation was observed between PL in the pregnancy orders. Using matrix H reduced the posterior standard deviation of the genetic correlations between PL and other traits. The genetic correlation between PL in heifers and STAY increased with the use of matrix H. Moderate to high and positive genetic correlations were observed between PL and STAY in primiparous and multiparous cows. High and moderate positive genetic correlations were observed between AFC and PL. Negative genetic correlation was observed between PL and SC450 or SC550. The genetic correlation between PL and ACP was generally medium to high and negative, with the greatest change in heifers when using matrix H. Medium or low genetic correlation was always observed between PL and W450 or W550. Including genomic information increased heritability for PL in heifers and improved the accuracy of genetic correlation estimates between PL and commonly evaluated traits. The third specific objective was to test the use of a reaction norm model (RNM) to identify GxE affecting PL in Brahman cattle, evaluating the inclusion of genomic information on VCE, predictive ability, and accuracy of GEBV. PL records in heifers were used along with meteorological data collected from NASAPOWER. The daily temperature and humidity index (THI) was calculated, and the date of pregnancy confirmation was used to associate PL records with the accumulated THI 30 days before and after confirmation. Threshold RNM were fitted for PL using matrices A and H. The use of matrix H increased the heritability of the regression intercept. The average heritability for PL on the accumulated THI gradient was similar using matrix A or matrix H. GxE affecting PL was confirmed by the negative genetic correlation between coefficients and between extreme values of accumulated THI. A significant reclassification between THI levels was observed because of GxE. VCE and genetic parameters changed with the use of matrix H, although no clear advantage was observed. However, the use of matrix H allowed more accurate and robust GEBV for random coefficients. There is evidence of GxE affecting the genetic response for PL in Brahman cattle under tropical conditions.

Resumo (português)

O ssGBLUP combina informações genômicas e de pedigree, o que pode aumentar a precisão na estimativa de componentes de variância (VCE). Em casos como os das características tradicionalmente avaliadas em programas comerciais de melhoramento genético de gado Nelore em Brasil, o uso do ssGBLUP frequentemente envolve a mistura de informações de um antigo programa de seleção baseado em pedigree com informações atuais da seleção genômica (GS). Isso é controverso porque as decisões de seleção são baseadas em informações genômicas, e apenas subconjuntos de animais são genotipados, o que representa um passo adicional de seleção que pode acrescentar viés nos valores genéticos genômicos (GEBV). Por outro lado, o ssGBLUP pode melhorar as VCE nos programas de GS para raças com menos registros, como é o caso do gado Brahman; principalmente nas aracterísticas com pouca herdabilidade, como a perda gestacional (PL), e quando considerada a interação genótipo – ambiente (GxE). Portanto, o objetivo geral foi avaliar a aplicação do ssGBLUP para VCE e predição genômica de características de produção e reprodução em duas populações de gado Zebu. No caso do gado Nelore, o primeiro objetivo específico foi estimar os componentes de variância para características de crescimento e reprodutivas utilizando duas diferentes matrizes de relacionamento (matriz A de relacionamento baseado no pedigree e a matriz H de relacionamento por pedigree e genômico), assim como considerando registros fenotípicos coletados antes ou depois da implementação da GS no gado Nelore Brasileiro. Além disso, avaliou-se como os GEBV são afetados pelas VCE e pela exclusão de registros anteriores à implementação da GS. Nesse caso, foram avaliadas as características peso aos 120 dias (W120), peso aos 450 dias (W450) e circunferência escrotal aos 450 dias (SC450). Foram utilizados três conjuntos de dados para a VCE: 1) incluindo todos os fenótipos (All), 2) apenas registros de animais nascidos antes de 2010 (before) ou 3) registros de animais nascidos desde 2010 (After), de acordo com o início da GS. Utilizamos a matriz A em todos os três conjuntos de dados (A_before, A_after e A_all), enquanto H foi usada em duas situações: H_All e H_After. Diferentes VCE foram usadas para a predição dos GEBV através do ssGBLUP. Esta etapa utilizou dois possíveis conjuntos de dados: todos os registros disponíveis (Dataset 1) e apenas registros coletados desde 2010 (Dataset 2). A validação foi conduzida com método LR. Para as três características, a herdabilidade aumentou usando A_After, e para W450 e SC450 a herdabilidade diminuiu com o uso da matriz H. Para W120, o Dataset 1 e VCE de A_After mostraram a maior precisão para o GEBV direto e maternos. Para W450 e SC450, o Dataset 1 e VCE usando H_After permitiram alta acurácia para os GEBV. Além disso, em ambas as características, usar o Dataset 2 não acrescentou viés nem dispersão e manteve alta a acurácia dos GEBV. Para W450 e SC450, a variância genética e a herdabilidade aumentaram com a GS. Para W120, os GEBV foram mais precisos usando A para a VCE. Foi possível obter alta acurácia dos GEBV para W450 e SC450 usando H na VCE e descartando registros anteriores à GS. É possível excluir registros anteriores à implementação da GS sem aumentar o viés e dispersão das predições. No caso do gado Brahman, o segundo objetivo específico foi estimar os componentes de variância e os parâmetros genéticos para PL e avaliar a sua correlação genética com características de crescimento e reprodutivas avaliadas em um programa de comercial de melhoramento genético, comparando o uso das matrizes A e H. Para isso, registros de PL de três ordens de gravidez (novilhas, vacas primíparas e vacas multíparas) foram coletados para dois rebanhos na Bolívia. Modelos de limiar uni-variados e multivariados foram ajustados para PL nas três ordens de gravidez, também se ajustaram modelos multivariados entre PL e características tradicionais como pesos ajustados aos 450 (W450) e 550 (W550) dias de idade, circunferência escrotal ajustada aos 450 (SC450) e 550 (SC550), produtividade acumulada da vaca (ACP), idade ao primeiro parto (AFC) e permanência (STAY). As matrizes A e H foram testadas em todos os modelos. A herdabilidade para PL foi baixa nas três ordens, mas o uso da matriz H aumentou a herdabilidade em novilhas. Observou-se alta correlação genética entre PL nas ordens de gravidez. Usar a matriz H reduziu o desvio padrão posterior das correlações genéticas entre PL e as demais características. A correlação genética entre PL em novilhas e STAY aumentou com o uso da matriz H. Correlações genéticas moderadas a altas e positivas foram observadas entre PL e STAY em vacas primíparas e multíparas. Correlações genéticas altas e médias e positivas foram observadas para AFC e PL. Correlação genética negativa foi observada entre PL e SC450 ou SC550. A correlação genética entre PL e ACP em geral foi média a alta e negativa, com maior mudança em novilhas quando usada a matriz H. Sempre foi observada correlação genética média ou baixa entre PL e W450 ou W550. Incluir informações genômicas aumentou a herdabilidade para PL em novilhas e melhorou a precisão das estimativas de correlação genética entre PL e características normalmente avaliadas. O terceiro objetivo específico foi testar o uso de um modelo de norma de reação (RNM) para identificar GxE que afeta a PL no gado Brahman, avaliando a inclusão de informações genômicas sobre as VCE, a habilidade de predição e a acurácia dos GEBV. Os registros de PL em novilhas foram usados junto com dados meteorológicos coletados do NASAPOWER. O índice de temperatura e umidade (THI) diário foi calculado e a data de confirmação da prenhez foi usada para associar os registros de PL com o THI acumulado 30 dias antes e depois da confirmação. RNM de limiar foram ajustados para PL usando as matrizes A e H. O uso da matriz H aumentou a herdabilidade do intercepto da regressão. A herdabilidade média para PL no gradiente de THI acumulado foi semelhante usando a matriz A ou matriz H. GxE afetando a PL foi confirmada pela correlação genética negativa entre os coeficientes e entre valores extremos de THI acumulado. Uma importante reclassificação entre níveis de THI foi observada como consequência da GxE. VCE e parâmetros genéticos mudaram com o uso da matriz H, embora nenhuma vantagem clara tenha sido observada. No entanto, o uso da matriz H permitiu GEBVs mas acurados e robustos para coeficientes aleatórios. Há evidências de GxE que afeta a resposta genética para PL em gado Brahman sob condições tropicais.

Descrição

Palavras-chave

Genômica, Melhoramento genetico, Bovinos de corte

Idioma

Inglês

Como citar

CARDONACIFUENTES, D. - Aplicação do ssGBLUP para estimativa de componentes de variância e predição genômica em duas populações de gado zebu - 2024, 106f - Tese (Doutorado em Genética e Melhoramento Animal) - Universidade Estadual Paulista, Jaboticabal, 2024.

Itens relacionados

Unidades

Departamentos

Cursos de graduação

Programas de pós-graduação