Modelo QSAR para previsão de potencial cancerígeno de compostos fenólicos em roedores através de rede neural artificial

• Ano de defesa: 2015
• Autor(a) principal: Vitorino Júnior, José
• Tipo de documento: Dissertação
• Tipo de acesso: Acesso aberto
• Idioma: por
• Instituição de defesa: Universidade Federal de Mato Grosso; Brasil; Instituto de Ciências Exatas e da Terra (ICET) – Araguaia; UFMT CUA - Araguaia; Programa de Pós-Graduação em Ciência de Materiais
• Palavras-chave em Português: Câncer; Compostos fenólicos; Rede neural; QSAR; CNPQ::CIENCIAS EXATAS E DA TERRA; Cancer; Phenols; ANN
• Link de acesso: http://ri.ufmt.br/handle/1/1455

The aim of the present study was the development of quantitative structureactivity relationship (QSAR) model to predict carcinogenic potency in phenolics compounds. The model was developed exploring the relationship between the experimental and predicted carcinogenic potency expressed as a tumorgenic dose TD50 for rats. A dataset of 65 phenolics substances was obtained after a preliminary screening of findings of rodent carcinogenicity for 1.547 chemicals accessible via Distributed Structure-Searchable Toxicity (DSSTOX) Public database network originated from the Lois Gold Carcinogenic Potency Database(CPDB). The Compounds chemical structures were optimization with Ghemical software and molecular descriptors were calculated using E-Dragon. The selection of the best descriptors was performed using multi filter with Genetic Search and Best First method and a new classification model were developed using the Radial Basis Function Neural Network in WEKA software. For the model validity we used a series of statistical indexes. We found that the model have good accuracy for the training set (74%). The model obtained a good predicted performance for the test set. It was obtained a good accuracy (88%), sensitivity (86%), and specificity (90%). A measure of model performance is provided also by the are a under the ROC curve. For training and test sets we have got a result of 0.886 and 0.927, respectively, which is promising initial result in modeling carcinogenicity.