Submitted by Marco Antônio de Ramos Chagas (mchagas@ufv.br) on 2018-08-31T14:31:04Z No. of bitstreams: 1 texto completo.pdf: 602062 bytes, checksum: 91584ea659213ec53e7d2c38ef94f83e (MD5) / Made available in DSpace on 2018-08-31T14:31:04Z (GMT). No. of bitstreams: 1 texto completo.pdf: 602062 bytes, checksum: 91584ea659213ec53e7d2c38ef94f83e (MD5) Previous issue date: 2018-07-04 / Conselho Nacional de Desenvolvimento Científico e Tecnológico / A seleção simultânea de caracteres, integrada à seleção genômica ampla (GWS), tem se tornado uma estratégia de grande interesse para os programas de melhoramento de plantas. Neste sentido, os objetivos desse estudo foram: i) comparar a acurácia e eficiência de seleção do método Multivariate Partial Least Square (Mpls) em relação aos métodos univariados: Random Regression Best Linear Unbiased Predictor (RRblup), Bayesian Lasso (Blasso) e Univariate Partial Least Square (Upls); ii) verificar a eficiência da seleção direta e indireta na GWS; iii) elaborar e comparar diferentes estratégias multivariadas, via índices de seleção integrados à GWS, eficientes na identificação e seleção precoce de indivíduos geneticamente superiores, em diferentes características simultaneamente. Dez populações F 2 com 800 indivíduos foram simuladas considerando quatro características com diferentes herdabilidades. Na primeira etapa da pesquisa, os dados simulados foram submetidos às análises de GWS via RRblup, Blasso, Upls e Mpls. Quatro índices de seleção genômica foram elaborados pelo somatório dos efeitos dos marcadores obtidos para cada característica, ponderados pela sua respectiva variância residual, sendo elaborado um índice para cada metodologia avaliada. Na segunda etapa da pesquisa, os dados simulados foram submetidos as análises de GWS via RRblup e MPls. Foram elaboradas e comparadas diferentes estratégias de índices de seleção aplicados à GWS: i) ponderar os efeitos dos marcadores pela variância residual (elaborado na primeira etapa); ii) codificar e padronizar os efeitos dos marcadores; iii) aplicar média nos efeitos dos marcadores; iv) aplicar o índice de Mulamba e Mock (1978) nos valores genéticos genômicos; v) codificar e padronizar os valores fenotípicos, antes das análises de GWS. Além disso, na segunda etapa dessa pesquisa, foram considerados dois cenários de seleção. No primeiro cenário, foram selecionados os indivíduos com maiores valores fenotípicos, valores genéticos verdadeiros e valores genéticos genômicos para as quatro características avaliadas. Já no segundo cenário, foi considerado diferente sentido de seleção para uma das características simuladas. As comparações entre os métodos e os índices de seleção foram realizadas considerando o tempo de processamento, as acurácias de predição, os ganhos de seleção e os coeficientes de coincidência de seleção. Foi aplicado o índice de Mulamba e Mock (1978) nos valores fenotípicos e valores genéticos verdadeiros. Os métodos de seleção genômica foram mais eficientes que a seleção fenotípica. O método Mpls foi similar ao método Upls para as características de menores herdabilidades e foi menos eficiente quando comparado aos métodos RRblup e Blasso. A seleção direta e indireta baseada nos valores genéticos genômicos foi mais eficiente que a seleção fenotípica. Nenhuma das estratégias avaliadas foi eficiente considerando diferente sentido de seleção para uma das características simuladas. Os índices ponderados pela variância residual apresentaram alta eficiência para aplicação na GWS, no entanto tenderam a maximizar os ganhos para as características de maiores herdabilidades. As estratégias de aplicar índices, via RRblup, a partir da média dos efeitos dos marcadores, dos valores fenotípicos codificados e padronizados e da aplicação do índice de Mulamba e Mock nos valores genéticos genômicos, resultaram em altos ganhos de seleção e mais se aproximaram aos ganhos obtidos pelo índice de Mulamba e Mock aplicado aos valores genéticos verdadeiros. A estratégia de codificar e padronizar os efeitos dos marcadores proporcionou os menores ganhos genéticos totais. De modo geral, os índices de seleção genômica propostos, proporcionaram maior eficiência de seleção, quando comparados ao índice de seleção de Mulamba e Mock fenotípico. Esses resultados sugerem que as estratégias multivariadas, via índices de seleção integrados à GWS, são promissoras para aplicação em programas de melhoramento genético de plantas, visando a seleção precoce direta de várias características simultaneamente. / Simultaneous traits selection, integrated with genome wide selection (GWS), has become a strategy of great interest for plant breeding programs. In this sense, the objectives of this study were: i) to compare the accuracy and efficiency of the Multivariate Partial Least Square (Mpls) method in relation to univariate methods: Random Regression Best Linear Unbiased Predictor (RRblup), Bayesian Lasso (Blasso) and Univariate Partial Least Square (Upls); ii) verify the efficiency of direct and indirect selection in GWS; iii) to elaborate and compare different multivariate strategies, through selection indexes integrated to GWS, efficient in the identification and early selection of genetically superior individuals, in several traits simultaneously. Ten F 2 populations with 800 individuals were simulated considering four traits with different heritabilities. In the first research step, the simulated data were submitted to the GWS analysis via RRblup, Blasso, Upls and Mpls. Four GWS indexes were elaborated by the sum of the markers effects obtained for each trait, weighted by their respective residual variance, and was elaborated an index for each methodology. In the second research step, the simulated data were submitted to GWS analysis via RRblup and MPls. Different selection index strategies applied to GWS were elaborated and compared: i) to weigh the effects of the markers by the residual variance (elaborated in the first step); ii) to encode and standardize the effects of markers; iii) to apply average markers effects; iv) to apply the Mulamba and Mock index (1978) to genomic breeding values; v) to encode and standardize phenotypic values, before to GWS analysis. In addition, in the second research step, two scenarios selection were considered. In the first scenario, individuals with higher phenotypic values, genetic values and genomic breeding values were selected for the four traits evaluated. In the second scenario, a different sense of selection was considered for one of the simulated traits. The comparisons among the methods and the selection indexes were performed considering the processing time, the prediction accuracy, the selection gains and the selection coincidence coefficients. The Mulamba and Mock index was applied to the phenotypic values and genetic values. The GWS methods were more efficient than phenotypic selection. The Mpls method was similar to the Upls method for the smaller heritabilities traits and was less efficient when compared to the RRblup and Blasso methods. The direct and indirect selection based on genomic breeding values was more efficient than phenotypic selection. None of the evaluated strategies was efficient considering a different sense of selection for one of the simulated traits. The residual variance weighted indexes showed high application efficiency to GWS, but tended to maximize gains for the traits of higher heritabilities. The strategies of applying indexes, via RRblup, from the markers effects mean, the coded and standardized phenotypic values and the application of the Mulamba and Mock index on the genomic breeding values, resulted in high selection gains and more approached to the gains obtained by the Mulamba and Mock index applied to the genetic values. The coding and standardizing strategy of the effects of markers provided the lowest total genetic gains. In general, the genomic selection indexes, provided greater selection efficiency than Mulamba and Mock selection index phenotypic. These results suggest that multivariate strategies, via selection indexes integrated to the GWS, are promising for application in plant genetic improvement programs, aiming at the direct early selection of several traits simultaneously.

Orientador: Luiz Carlos Forti / Banca: Odair Correa Bueno / Banca: Ivone Paschoal Garcia / Abstract: The objective of this paper was to have a better understanding of plant-ant-symbiotic fungus interaction focusing on foraging behavior of Atta sexdens rubropilosa workers during plant selection. Physical and chemical substrate factors were approached which can have a role on foraging material selection for symbiotic fungus cultivation besides defoliating localization in artificial plants. Tested hypothesis was that besides chemical characteristics physical stimulus and leaves localization in a plant also play a very important role in substrate selection by A. sexdens rubropilosa workers. Different shape and thickness material was used (different leaves, paper leaves with different shapes and thickness impregnated in plant extract and artificial plants). The aim was to verify the existence of physical and chemical resistance, leaf palatability and defoliating localization. Wood plants Actinostemon communis, Alchornea triplinervea, Croton floribundus, Faramea cyanea, were offered to workers and evaluated 4 according to mechanical resistance of cut and palatability. Each plant was offered individually in big disc shape (2,5cm diameter), small disc shape (0,5cm diameter) and whole leaves. Other studies were carried out for detecting physical and chemical stimulus through simulated cuts and impregnation of plant extract in paper leaves with different thickness making it possible to evaluate the combinations between physical and chemical substrate characteristics. Defoliating localization in plants was studied when offering artificial plants with four levels where either wood plant or Ligustrum sp. leaves was attached. Defoliating intensity was measured by the number of fallen leaves by ants. In order to study wood plant attractiveness small disc shaped leaves (0,5cm diameter) were offered at the same time for ants in laboratory. The end of the experiment was determined either by the carrying... (Complete abstract, click electronics address below). / Mestre

Made available in DSpace on 2014-06-11T19:28:36Z (GMT). No. of bitstreams: 0 Previous issue date: 2006-02-10Bitstream added on 2014-06-13T20:37:43Z : No. of bitstreams: 1 noronhajunior_nc_me_botfca.pdf: 814511 bytes, checksum: f0ba517a59245727f01aed36f56871ba (MD5) / Conselho Nacional de Desenvolvimento Científico e Tecnológico (CNPq) / Coordenação de Aperfeiçoamento de Pessoal de Nível Superior (CAPES) / Empresa Privada / The objective of this paper was to have a better understanding of plant-ant-symbiotic fungus interaction focusing on foraging behavior of Atta sexdens rubropilosa workers during plant selection. Physical and chemical substrate factors were approached which can have a role on foraging material selection for symbiotic fungus cultivation besides defoliating localization in artificial plants. Tested hypothesis was that besides chemical characteristics physical stimulus and leaves localization in a plant also play a very important role in substrate selection by A. sexdens rubropilosa workers. Different shape and thickness material was used (different leaves, paper leaves with different shapes and thickness impregnated in plant extract and artificial plants). The aim was to verify the existence of physical and chemical resistance, leaf palatability and defoliating localization. Wood plants Actinostemon communis, Alchornea triplinervea, Croton floribundus, Faramea cyanea, were offered to workers and evaluated 4 according to mechanical resistance of cut and palatability. Each plant was offered individually in big disc shape (2,5cm diameter), small disc shape (0,5cm diameter) and whole leaves. Other studies were carried out for detecting physical and chemical stimulus through simulated cuts and impregnation of plant extract in paper leaves with different thickness making it possible to evaluate the combinations between physical and chemical substrate characteristics. Defoliating localization in plants was studied when offering artificial plants with four levels where either wood plant or Ligustrum sp. leaves was attached. Defoliating intensity was measured by the number of fallen leaves by ants. In order to study wood plant attractiveness small disc shaped leaves (0,5cm diameter) were offered at the same time for ants in laboratory. The end of the experiment was determined either by the carrying... (Complete abstract, click electronics address below).

Submitted by Marco Antônio de Ramos Chagas (mchagas@ufv.br) on 2017-02-13T15:49:16Z No. of bitstreams: 1 texto completo.pdf: 3558878 bytes, checksum: 2b30ec8d9f71e9e1bab44b69484c0c54 (MD5) / Made available in DSpace on 2017-02-13T15:49:16Z (GMT). No. of bitstreams: 1 texto completo.pdf: 3558878 bytes, checksum: 2b30ec8d9f71e9e1bab44b69484c0c54 (MD5) Previous issue date: 2016-11-30 / Conselho Nacional de Desenvolvimento Científico e Tecnológico / A seleção genômica (SG) tem se tormado uma ferramenta de grande potencial no melhoramento de plantas. Além dela, o estudo de associação genômica (EAGA) e a seleção assistida por marcadores moleculares (SAM) também são metodologias com aplicabilidade no melhoramento. A diferença básica entre essas metodologias é que enquanto a SAM utiliza mapas de ligação e o EAGA utiliza mapas de associação para identificar marcadores significativos, a SG utiliza todos os marcadores disponíveis sem a necessidade de nenhum tipo de mapa. Portanto os objetivos desta pesquisa foram: 1) avaliar modelos utilizando os SNPs significativos encontrados pelos SAM e EAGA como efeito fixo nos modelos comumente utilizados na SG, em que no modelo tradicional, todos os SNPs são estabelecidos como de efeito aleatório. Estes modelos foram comparados com o modelo padrão utilizado na SG (RRBLUP bayesiano); 2) comparar os métodos tradicionais de seleção genômica (todos os SNPs como efeito aleatório); 3) verificar como a herdabilidade e o número de QTLs que controlam a característica podem influenciar na predição do valor genético; 4) estabelecer uma equação de predição da correlação genética em função da correlação fenotípica; 5) estabelecer o número ideal de indivíduos para compor a população de treinamento e; 6) estabelecer a quantidade necessária de marcadores para obter máxima acurácia pelos métodos de seleção genômica. Foram simuladas populações F 2 com 1.000 indivíduos em diferentes cenários. As populações foram simuladas com 4.500 (objetivo 1) e 3.000 marcadores (demais objetivos). Foram simuladas características com diferentes herdabilidades (5, 20, 40, 60, 80 e 99%) e o número de QTLs (60, 120, 180 e 240) (objetivos 2, 3 e 4). Foram estimados para todos os cenários a capacidade preditiva fenotípica e genotípica, a acurácia fenotipica e genotípica, a herdabilidade genômica, a variância genética, o ganho com a seleção, o índice de coincidência e o tempo de processamento. Foi utilizado a cross validação 5-fold com 50 repetições. As principais conclusões desta pesquisa foram: 1) A utilização de um modelo de SG com as marcas significativas encontradas pelo EAGA como efeito fixo e as demais marcas como efeito aleatório é uma boa estratégia para selecionar indivíduos superiores com alta acurácia; 2) A introdução no modelo de SG de QTLs que já foram descritos previamente para a característica em estudo, como efeito fixo, permite a seleção de indivíduos superiores de forma mais acurada; 3) os modelos de seleção genômica para predição em populações F 2 devem ser compostos por 200 a 900 marcadores de maior efeito sobre a característica e mais de 600 indivíduos na população de treinamento. / Genomic selection (GS) has become a high potential tool in plant breeding. Moreover, genomic wide association study (GWAS) and marker-assisted selection (MAS) are also methodologies with great potential in plant breeding. The basic difference among them is while MAS requires linkage mapping and GWAS requires association mapping to identify significant markers, GWS performs all available markers without any mapping. Therefore, the objectives in this research were: 1) to evaluate models using significant SNPs found by GWAS and MAS as fixed effect in the widely GS models, which, in the traditional model all SNPs are treated as random effect. These models were compared with the standart GS model (Bayesian RRBLUP); 2) To compare the most GS traditional models (all SNPs as random effect); 3) to verify how the heritability and number of QTLs which control a specific trait can influence for predicting genetic value; 4) to establish a prediction equation to estimate the genetic correlation based on phenotypic correlation; 5) to establish the optimal number of individuals to compose the training population and; 6) to establish the number of markers needed to obtain the maximum accuracy by the genomic selection methods. F 2 population was simulated with 1,000 individuals in several scenarios. Populations were simulated with 4,500 (objective 1) and 3,000 markers (other objectives). Traits with different heritability (5%, 20%, 40%, 60%, 80% and 99%) and numbers of QTLs (60, 120, 180 and 240 – objectives 2, 3, and 4) were simulated. Phenotypic and genotypic predictive ability, phenotypic and genotypic accuracy, genomic heritability, genetic variance, selection gain, conincidence index, and processing time were estimated for all scenarios. 5-fold cross validation was repeated 50 times. The mainly conclusion in this research were: 1) SG model performed with the significant markers found by GWAS as fixed effect and the remaining SNPs as random effects is a useful strategy to select superior individuals with high accuracy; 2) GS model performed with the QTLs, previously reported for the traits in study, as fixed effect allows the selection of superior individuals more accurate; 3) Genomic selection models should be composed with number of markers ranged from 200 to 900 and number of individuals in the training population beyond 600.