Aplica??o de modelos ARTMAP na predi??o de resist?ncia do HIV-1 aos Inibidores de protease

• Ano de defesa: 2014
• Autor(a) principal: Calin, Alessandro dos Santos
• Orientador(a): Silva, Robson Mariano da
• Banca de defesa: Silva, Robson Mariano da, Oliveira, Francisco Bruno Souza, Delgado, Angel Ramon Sanchez
• Tipo de documento: Dissertação
• Tipo de acesso: Acesso aberto
• Idioma: por
• Instituição de defesa: Universidade Federal Rural do Rio de Janeiro
• Programa de Pós-Graduação: Programa de P?s-Gradua??o em Modelagem Matem?tica e Computacional
• Departamento: Instituto de Ci?ncias Exatas
• País: Brasil
• Palavras-chave em Português: HIV-1; Redes Neurais Artificiais (RNAs); Terapia Antirretroviral
• Palavras-chave em Inglês: Artificial Neural Networks (ANN); Antiretroviral Therapy
• Área do conhecimento CNPq: Matem?tica
• Link de acesso: https://tede.ufrrj.br/jspui/handle/jspui/3338

Through the last decade, the Antiretroviral Therapies (ART) reduced the mortality of patients with HIV-1. Nevertheless, this decrease could not completely prevent the emergence of new resistant viral forms, mainly caused by the high mutational rate of HIV-1. The development of resistance of HIV-1 to antiretroviral (ARV) is a limiting factor for the success of ART. Because, in addition to not respond adequately to treatment, patients with resistant virus can transmit these mutant viruses, representing a serious public health problem. Patients with virological failure, usually, require changes to their antiretroviral regimens, thus, techniques that can assist in the prediction of resistance to antiretroviral therapies allow to minimize the gaps and, consequently, prevents the increase of the viral load of patients. In view of these facts, we developed this study with the goal of developing two computational models: one based on ArtMap Neural Networks and other Fuzzy ArtMap Neural Networks. In order to investigate resistance to antiretroviral therapy for HIV-1 protease inhibitors (IPs) for subtypes B and C. To apply the methodology we use the data obtained from the Laboratory of Molecular Virology at Universidade Federal do Rio de Janeiro (UFRJ, Brazil) and a public base courtesy of Stanford University (SU, United States). Prior to divide the test set (30%) and training (70%), was made a preprocessing analyzing the frequency of occurrence of mutations in all positions of the protease. And it was found that some positions had very low mutation rates, which would have significance for the categorization of samples in the general database, and thus, was considered for the model only the positions of a total equal to or greater than 7.5% of mutations. The results were significant in both models, especially in groupings to patients resistant to Lopinavir, Nelfinavir and the non-resistant antiretroviral patients. The results were made using the concept of specificity, sensitivity and accuracy..