Métodos de avaliação da eficiência da multiplicação genética por produção de embriões bovinos

• Ano de defesa: 2014
• Autor(a) principal: Penitente Filho, Jurandy Mauro
• Orientador(a): Torres, Ciro Alexandre Alves
• Banca de defesa: Santos, Giancarlo Magalhães dos , Guimarães, José Domingos
• Tipo de documento: Tese
• Tipo de acesso: Acesso aberto
• Idioma: por
• Instituição de defesa: Universidade Federal de Viçosa
• Programa de Pós-Graduação: Doutorado em Zootecnia
• Departamento: Genética e Melhoramento de Animais Domésticos; Nutrição e Alimentação Animal; Pastagens e Forragicul
• País: BR
• Palavras-chave em Português: Bovino - Reprodução; Fertilização in vitro; Redes neurais artificiais
• Palavras-chave em Inglês: Cattle - Breeding; IVF; Artificial neural networks
• Área do conhecimento CNPq: CNPQ::CIENCIAS AGRARIAS::MEDICINA VETERINARIA::REPRODUCAO ANIMAL
• Link de acesso: http://locus.ufv.br/handle/123456789/1878

The objective of the present study was to evaluate some strategies for increasing the efficiency of bovine in vitro embryo production (IVP). Two articles were developed, first one aimed to evaluate whether ovarian volume, presence and diameter of the CL have effect on the number and quality of bovine recovered oocytes. The second article aimed to compare different statistical methods to predict pregnancy in a commercial program of IVP. Article 1: one hundred and ten ovaries were obtained from slaughterhouse. Cumulus oocytes complex were aspirated and evaluated under stereomicroscope. Oocytes were counted and classified according to their quality (Grade I, II, III and IV). Ovarian volume was weakly correlated to the number of good quality oocytes (P < 0.05). Ovaries with CL show greater number of good quality oocytes than ovaries without CL (P < 0.05). Besides, presence of CL and its diameter positively influenced the probability of recovering good quality oocytes (P < 0.05). In conclusion, ovarian volume is not a good parameter itself to predict important ovarian characteristics; moreover, analysis of CL, its presence and diameter, may be a good tool to improve efficiency on IVP programs. Article 2: Artificial neural networks (ANN) and decision trees have proved to be very successful in different fields of study such as medicine, genetics and animal production. However, the use of these methodologies in bovine reproduction is still very scarce. In this study, a real dataset was used, consisting of 9,697 embryo transfers, 6,788 observations were used in training model and 2,909 were used for validation. Dataset was analyzed by logistic regression, feed-forward neural network, multilayer perceptron with one and two hidden layers, and a based on ID3 algorithm decision tree. All analyses were compared by ROC curve, predictive capacity, Kolmogorov-Smirnov test, and point biserial correlation. In conclusion, all analyzes were very similar. Nevertheless, analysis by the feed-forward neural network correctly classified more than 70 % of positive pregnancies. Despite having incorrectly classified almost 60 % of non-pregnancies, this result becomes interesting because not detect a positive pregnancy is more harmful to a commercial program than not detect a negative pregnancy. Still, studies with larger dataset and more explanatory variables are required, including not only categorical variables, but also quantitative variables to ensure more accurate results.