Divergência genética e predição de valores genotípicos em soja

• Autor(a) principal: Godoi, Cláudio Roberto Cardoso de
• Data de Publicação: 2014
• Tipo de documento: Tese
• Idioma: por
• Título da fonte: Repositório Institucional da UFG
• Texto Completo: http://repositorio.bc.ufg.br/tede/handle/tede/3884
• topic: Glycine max; SSR; Introdução de germoplasma; Variabilidade genética; REML/BLUP; Matriz de parentesco; Modelo misto; Modelo fixo; Glycine max; SSR; Germplasm introduction; Genetic variability; REML/BLUP; Relationship matrix; Mixed model; Fixed model; GENETICA::GENETICA VEGETAL

Soybean breeding programs practice selection of high genetic value genotypes with two main objectives: a) to use them as parents in the hybridization process (first stage of the program), and b) to indicate them as new cultivars (final stage of the program). In this context, a first study used microsatellite markers (SSR) to assess the genetic diversity of soybean germplasm adapted to the Brazilian conditions. The experimental material consisted of 192 accessions, which included both introductions and Brazilian germplasm. The genetic divergence was assessed by descriptive analysis and the Rogers-W genetic distance. A total of 222 alleles were identified in the 37 genotyped loci, with an average of six alleles and a range of 2 to 14 alleles per locus. The genotypes were clustered according to the origin of the germplasm, and resulted in two groups: one group formed by introductions and other by Brazilian genotypes. Eighty five percent of the genetic distances estimates were above 0.70, suggesting that the assessed germplasm has good potential for hybridization in soybean breeding programs. It was concluded that the SSR markers are useful to identify divergent genotypic groups, as well as genotypic combinations with high genetic variability. It also became clear that the use of introduced germplasm ensures the incorporation of alleles necessary to increase the genetic base of soybeans and, consequently, the variability needed for the selective process. In a second study, the mixed model approach was used to assess some strategies of estimation and prediction of genotypic values for grain yield in the soybean regional yield trials. A total of 111 genotypes classified into three maturity groups were sown in up to 23 experiments in Central Brazil. The experiments were carried out in randomized complete block designs, with three replications. The biometrical analyses followed the fixed model and mixed model approaches, in the latter case assuming the genotypic effects as random. In the mixed model approach, analyses were made with or without information from the relationship estimates obtained either by genealogy or SSR markers, arranged in a genotypic covariance matrix (G). Also, in a context of spatial analysis, different structures were used in the residual covariance matrix (R) for each mixed model adjusted. The following conclusions were obtained: i) the fixed model analysis is adequate to estimate genotypic values in soybean trials with balanced data and orthogonal design; ii) under such conditions and intermediate to low heritability, the inclusion of relationship information associated to G matrix, although does not ensure the best fit models, improves the precision in predicting genotypic values; iii) the use of spatial structures associated to R matrix, in presence of the residual autocorrelation, improves the goodness of model fit to the data; and, iv) the choice of model for the analysis does not change the ranking of the genotypes in high heritability situations and, therefore, does not impact significantly on the selection of superior genotypes.