A carregar...
Publicação
Modelos normal-assimétrico e alélico no mapeamento de QTLs
| Nome: | Descrição: | Tamanho: | Formato: | |
|---|---|---|---|---|
| 1.17 MB | Adobe PDF | |||
| 11.02 KB | Internal XML |
Resumo(s)
Esta dissertação aborda o mapeamento de loci genéticos que controlam caracteres quantitativos (quantitative trait loci - QTLs) em cruzamentos experimentais bi-alélicos. Existem vários métodos estatísticos de mapeamento de QTLs, tais como o mapeamento intervalar (IM) e o mapeamento intervalar composto (CIM) que assumem a normalidade do carácter quantitativo dado o genótipo do suposto QTL. No entanto, se a suposição de normalidade é violada, pode-se detectar um QTL inexistente ou o poder de detecção de QTLs pode diminuir. Nesta dissertação propõem-se três novos métodos estatísticos, aplicáveis no mapeamento genético de caracteres quantitativos que não apresentam distribuição normal. O primeiro método proposto assume que os alelos de um QTL actuam independentemente, sendo designado por mapeamento intervalar alélico (aIM). No aIM, a distribuição do carácter quantitativo é uma mistura de duas componentes, cada uma associada a um dos dois alelos do QTL, e as proporções de mistura são funções da penetrância alélica e da penetrância de factores externos. Pela análise de dois conjuntos de dados reais e dados simulados, concluiu-se que o método aIM tem maior poder de detecção de QTLs quando a penetrância alélica diminui ou a penetrância de factores externos aumenta. Em paralelo desenvolveu-se uma abordagem que supõe que o carácter quantitativo tem distribuição normal-assimétrica para cada classe genotípica do suposto QTL. Esta abordagem é implementada nos dois métodos referidos anteriormente, IM e CIM, resultando no mapeamento intervalar normal-assimétrico (snCIM) e no mapeamento intervalar composto normal assimétrico (snCIM), respectivamente. Ambos os métodos, foram aplicados a um conjunto de dados reais e foi realizado um estudo de simulação. O snIM e o snCIM revelaram maior poder de detecção de QTLs do que o IM ou o CIM, respectivamente, principalmente quando o carácter quantitativo apresenta uma distribuição com assimetria elevada. Esta dissertação demonstra a relevância do desenvolvimento de novos modelos genéticos e abordagens estatísticas para analisar dados que não verifiquem as suposições dos métodos vigentes de mapeamento de QTLs.
This thesis is dedicated to develop methodology for the mapping of genetic loci controlling quantitative traits (quantitative trait loci - QTLs) in bi-allelic experimental crosses. A number of QTL mapping methods have been proposed namely, the interval mapping (IM) and the composite interval mapping (CIM) that are based on a genetic model that assumes the normality of the phenotypic distribution given the genotype. Nevertheless, when normality is violated this assumption may lead to inaccurate results including detection of false QTLs or undetected QTLs. This thesis proposes three statistical methods to be applied to the genetic mapping of quantitative traits that do not show a normal distributed. First, it is proposed a mapping method based on a genetic model that considers independent allelic action, designed as allelic interval mapping (aIM). According to this method the phenotypic distribution is a mixture of two normal components, each associated to one QTL allele, where the mixture weights are dependent on allelic and external penetrance. Using two real data sets and simulations studies it is shown that aIM has higher power to detect QTLs when allelic penetrance is low or external penetrance is high. Next, is proposed analysis methodology that assumes the phenotype has skew-normal distribution in each genotypic class. This approach is implemented in two previously described QTL analysis methods, IM and CIM, giving rise to the skew-normal interval mapping (snIM) and the skew-normal composite interval mapping (snCIM), respectively. Both methods were applied to real data analysis and were submitted to simulation studies, having revealed higher detection power as compared to their original counterparts when the asymmetry of the phenotypic distribution was high. This thesis covers a subject that has been neglected in the field of QTL mapping and points out to the need of developing new genetic models and statistical approaches to treat data falling outside the current QTL mapping assumptions.
This thesis is dedicated to develop methodology for the mapping of genetic loci controlling quantitative traits (quantitative trait loci - QTLs) in bi-allelic experimental crosses. A number of QTL mapping methods have been proposed namely, the interval mapping (IM) and the composite interval mapping (CIM) that are based on a genetic model that assumes the normality of the phenotypic distribution given the genotype. Nevertheless, when normality is violated this assumption may lead to inaccurate results including detection of false QTLs or undetected QTLs. This thesis proposes three statistical methods to be applied to the genetic mapping of quantitative traits that do not show a normal distributed. First, it is proposed a mapping method based on a genetic model that considers independent allelic action, designed as allelic interval mapping (aIM). According to this method the phenotypic distribution is a mixture of two normal components, each associated to one QTL allele, where the mixture weights are dependent on allelic and external penetrance. Using two real data sets and simulations studies it is shown that aIM has higher power to detect QTLs when allelic penetrance is low or external penetrance is high. Next, is proposed analysis methodology that assumes the phenotype has skew-normal distribution in each genotypic class. This approach is implemented in two previously described QTL analysis methods, IM and CIM, giving rise to the skew-normal interval mapping (snIM) and the skew-normal composite interval mapping (snCIM), respectively. Both methods were applied to real data analysis and were submitted to simulation studies, having revealed higher detection power as compared to their original counterparts when the asymmetry of the phenotypic distribution was high. This thesis covers a subject that has been neglected in the field of QTL mapping and points out to the need of developing new genetic models and statistical approaches to treat data falling outside the current QTL mapping assumptions.
Descrição
Tese de doutoramento, Estatística e Investigação Operacional (Estatística Experimental e Análise de Dados), Universidade de Lisboa, Faculdade de Ciências, 2009
Palavras-chave
Estatística e investigação operacional Estatística experimental Análise de dados Teses de doutoramento - 2009