Seleção combinada, massal e entre e dentro, análise de trilha e repetibilidade em progênies de meios-irmãos de maracujazeiro (Passiflora edulis f. flavicarpa)

• Ano de defesa: 2006
• Autor(a) principal: Negreiros, Jacson Rondinelli da Silva
• Orientador(a): Bruckner, Claudio Horst
• Banca de defesa: Cruz, Cosme Damião , Motoike, Sérgio Yoshimitsu , Siqueira, Dalmo Lopes de , Soares, Plínio Cesar
• Tipo de documento: Tese
• Tipo de acesso: Acesso aberto
• Idioma: por
• Instituição de defesa: Universidade Federal de Viçosa
• Programa de Pós-Graduação: Doutorado em Genética e Melhoramento
• Departamento: Genética animal; Genética molecular e de microrganismos; Genética quantitativa; Genética vegetal; Me
• País: BR
• Palavras-chave em Português: Maracujá; Seleção; Passiflora edulis
• Palavras-chave em Inglês: Passion fruit; Selection; Passiflora edulis
• Área do conhecimento CNPq: CNPQ::CIENCIAS AGRARIAS::AGRONOMIA::FITOTECNIA::MELHORAMENTO VEGETAL
• Link de acesso: http://locus.ufv.br/handle/123456789/1295

The possibility to foresee selection gains is an important contribution of the quantitative genetics for the genetic improvement. The gain prediction helps breeders to choose the best selection method. The yield is normally an important trait to be improved. Fruit quality is a major aim in fruit breeding programs, while fruit quality is an important character for the market. The present work was carried out in the Department of Plant Science of the Federal University of Viçosa (UFV) and aimed to study the best selection criteria to select plants among and inside half-sib yellow passion fruit progenies. Also, to analyze the relationships among fruit traits and their direct and indirect effects and to estimate the coefficients of repeatability of the fruit traits. Thirty seven two years old yellow passion fruit half sib progenies were evaluated from January to May 2004. The experiment was outlined as a randomized block design, with 37 treatments, 3 replications and 4 plants per plot. The planting distances were 3.5 x 3.5 m. The plants were trained in 1.80 m high one-wire trellises. The evaluated traits were: length, diameter and mass of the fruit, mass of the pulp, mass and thickness of the skin, total soluble solids content, number of fruits per plant and estimated yield. Selection gains were estimated with the selection of 25% among and 50% inside the families, for each evaluated traits. All the traits were submitted to a positive selection, except skin weight and thickness, which need to be decreased. The selection criteria studied were: conventional selection among and inside families, index of combined selection, mass selection and stratified mass selection. The path coefficient were studies first associating the estimated yield as basic variable to their primary components: length and diameter of the fruit, skin thickness and total soluble solid contents. In the second causal diagram the estimated yield was associated to their primary components - number and weight of the fruit, and the secondary components - mass of the pulp and of the skin. The repeatability coefficients were estimated by the variance analysis methods (ANOVA), main components (CP), based on the matrixes of correlations and co-variances, and structural analysis (AE), based on correlations matrix. It was found that all the traits had genetic variability and all the selection criteria used were efficient for the improvement. The highest gain expectation was given by the combined selection index, which was considered the most appropriate selection method for the studied population. In the path coefficient analysis, the second causal model explained better the variability on the yield, where the number of fruits had higher direct effect. The mass of the pulp was the most important secondary character for the yield. The repeatability analysis showed that length and mass of fruits, mass of pulp and mass of skin were highly regular from one season to another. Differences were verified among the repeatability coefficients estimated by the method of ANOVA and the multivariate methods. The main components method based on the co-variance matrix gave the highest estimatives.