Modelo de fragilidade gama e regressão quantílica em análise de sobrevivência de abelhas melíferas expostas à proteína Cry1Ac

• Ano de defesa: 2014
• Autor(a) principal: Samudio, Fanni Petrona Ruiz
• Orientador(a): Martins Filho, Sebastião
• Banca de defesa: Nascimento, Ana Carolina Campana , Silva, Fabyano Fonseca e
• Tipo de documento: Dissertação
• Tipo de acesso: Acesso aberto
• Idioma: por
• Instituição de defesa: Universidade Federal de Viçosa
• Programa de Pós-Graduação: Mestrado em Estatística Aplicada e Biometria
• Departamento: Estatística Aplicada e Biometria
• País: BR
• Palavras-chave em Português: Abelha - Criação; Abelhas melíferas; Apis mellifera; Proteína Cry 1Ac
• Palavras-chave em Inglês: Bee - Creation; Honeybees; Apis mellifera; Cry 1AC protein
• Área do conhecimento CNPq: CNPQ::CIENCIAS AGRARIAS
• Link de acesso: http://locus.ufv.br/handle/123456789/4077

Bees are essential for the maintenance of biodiversity. They are also responsible for a large percentage of world food production. The species Apis mellifera is considered of great economic value due to its products for human consumption. The creation of insect-resistant transgenic products has increased the chances for the contact between bees and the Cry proteins derived from the bacterium Bacillus thuringiensis, which can be toxic to bees. Therefore, it is very important to study the hazards of toxicity. The study of these proteins in bees can be conducted by using survival analysis techniques, in which the response variable is the time until the occurrence of the event of interest, called failure time. If failure does not occur, time is called censoring, which is some partial information. The main interest is to estimate parameters to describe the survival or hazards, at a given time. In general, when there are covariates that may affect the survival time, adjustment can be performed by the Cox Regression Model, also known as proportional hazards model, through the assumption of proportional hazards among individuals over time. Chapter I, which deals with an adaptation of the Cox Regression model known as Fragility Model, not only explains the hazard of failure from individuals due to the effect of covariates, but also describes the existence of an unobserved random variable that groups individuals in natural or artificial conglomerates. The survival times were modeled to explain the hazard of failure by immature workers of A. mellifera under the effect of covariates; the colony was used as a random variable; and the ingestion of Cry1Ac protein, as fixed explanatory variable. Three different diets were tested to evaluate the toxicity of Cry1Ac on A. Mellifera: pure artificial (D0), artificial diluted in water with Cry1Ac (D2) and artificial diluted in water (D2). The individuals were collected from five different colonies maintained in Viçosa, Minas Gerais, Brazil. The random variable colony (frailty) was significant, which indicates statistical differences in the lifespans of bees from different colonies. Considering this diversity of frailty, the artificial diet diluted in water showed a higher hazard of failure, significantly different from the effect of the control diet (artificial pure). Therefore, the survival of the bee larvae was decreased due to the addition of water to the diet, which diluted the feed. However, the Cry1Ac-protein-based diet showed no significant hazard of failure when compared with the control. An alternative technique to the Cox model is presented in Chapter II of this work. When no proportional hazards of the individuals over time are observed within the sample, it is necessary to stratify or perform some other weakening of the proportional hazard condition. It can be used the Quantile Regression, which studies the relationship between the dependent variable and the explanatory variables in the conditional quantiles by minimizing the weighted mean absolute errors. This technique has properties of equivariance, invariance for monotomic transformations and robustness in the presence of outliers. Thus, by means of equivariance, they can be applied to data with censoring. It was found that, in the same set of biosafety data from the Cry1Ac protein, in honeybees called A. Mellifera, the hazards of failure by individuals are not proportionalities for the different diets studied. The survival times of bees were adjusted by quantile regression, by using the Portnoy estimator for 14 quantiles. In quantiles {0.10; 0.15; 0.30; 0.40}, the coefficients are negative values significantly different from zero. For such, in these quantiles, the individuals fed with pure diet (D0) had longer survival time compared to those fed with a diet containing Cry1Ac protein (D1). This was observed among younger individuals, since the lifespan of the larvae lifespans in these quantiles are below the average time of life. The coefficients for the quantiles {0.35; 0.50; 0.60} presented statistically significant negative effect. Therefore, water addition to the diet affected the survival of larvae approximately at the average lifespans.