Coleções nucleares e associação do teor de óleo de cártamo com variáveis ecogeográficas por inteligência computacional

• Ano de defesa: 2014
• Autor(a) principal: Boldt, Alberto Souza
• Orientador(a): Motoike, Sérgio Yoshimitsu
• Banca de defesa: Cecon, Paulo Roberto , Sá, Rogério Oliveira de , Dias, Luiz Antonio dos Santos
• 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: Plantas oleoginosas; Carthamus tinctorius; Açafrão; Germoplasma vegetal; Coleção nuclear
• Palavras-chave em Inglês: Oleaginous plants; Carthamus tinctorius; Saffron; Plant germplasm; Core collection
• Área do conhecimento CNPq: CNPQ::CIENCIAS AGRARIAS::AGRONOMIA::FITOTECNIA::MELHORAMENTO VEGETAL
• Link de acesso: http://locus.ufv.br/handle/123456789/1400

Safflower (Carthamus tinctorius L.) is an oilseed species with a large genetic potential available in genebanks. Source of relevant characters, safflower germplasm banks have shown limited use due to the large number of accessions available in collections. The present study aimed to explore the genetic diversity of safflower through the establishment of more expressive core collections using maximization strategy and the stratification of genotypes in genetic groups. The study also aimed to investigate the existence of predictive association between oil content and ecogeographic parameters of the original site of safflower accessions, using the focused identification of germplasm strategy to explore the association and increase the chances of finding safflower genotypes with high oil content. Core collections were established using phenotypic qualitative and quantitative traits data of 1640 safflower accessions from 48 countries. The accessions were stratified into genetic groups according to country s origin and sampled according to the maximization strategy (M strategy) . The established core collections were compared with the base collection using the following validation statistics: chi-square test, mean difference, difference of variances, coincidence rate, variable rate and Shannon index. Magnitude estimates of validation statistics indicated that base collection s genetic variability was preserved in the core collections based on safflower centers of similarity. Core collections stratified by genetic groups consisted in about 60 genotypes, with a mean difference of 7% over the base collection and coincidence rate above 94%. The combined use of the maximization strategy and stratification of genotypes in genetic groups maximized the capture of genetic variation and introduced more efficiency, establishing core collections with a fewer number of accessions. The core collections included approximately 3,75% accessions conserved in safflower base collection. To establish expressive core collections is necessary selecting accessions properly from base collection. The Focused Identification Germplasm Strategy (FIGS) is an efficient method to optimize the selection of useful accessions kept in collections. The FIGS makes use of predictive association between characteristics and environmental variables in the search for genotypes with high probability of containing the trait of interest. The present study aimed to investigate the existence of predictive association between oil content and ecogeographic parameters of the original site of safflower genotypes using the FIGS based on machine learning approaches. Random forests, support vector machines and artificial neural networks were used to model the association between oil content of 100 safflower genotypes and 56 ecogeographic parameters. The models accuracies indicated that the distribution of safflower genotypes with high oil content is not random but associated to environmental factors, even with some degree of overlap between the oil content in some environments. The final results suggest that exploring the predictive association between oil content and ecogeographic parameters of original collection site of safflower germplasm increases the chances of finding genotypes with high oil content.