Classificadores de regressão logística, Naive Bayes e Random Forest na análise do Tropismo do HIV-1 de subtipo B

• Ano de defesa: 2019
• Autor(a) principal: Barros, Cesar Borges
• Tipo de documento: Dissertação
• Tipo de acesso: Acesso aberto
• Idioma: por
• Instituição de defesa: Universidade Federal do Rio de Janeiro; Brasil; Instituto Alberto Luiz Coimbra de Pós-Graduação e Pesquisa de Engenharia; Programa de Pós-Graduação em Engenharia Biomédica; UFRJ
• Palavras-chave em Português: HIV-1 subtipo B; Tropismo viral; Classificadores baseados em aprendizado de máquina; CNPQ::ENGENHARIAS
• Link de acesso: http://hdl.handle.net/11422/13240

The development of coreceptor antagonists – such as maraviroc – for HIV treatment has made mandatory the clinical determination of viral coreceptor usage prior to rescue therapy. Technical issues presented by TrofileTM, the gold standard phenotypic assay, hindered its use as a routine diagnostic tool. This fact has lead to the development of genotypic algorithms, whose evaluations are based on DNA sequences of the V3 region of HIV-1 gp120. These algorithms proved to be cheaper, easier to use, and less time consuming than the phenotypic method. One of them, geno2pheno has also gained widespread use since it showed 86.5% predictive concordance with TrofileTM. The present project aimed to develop accurate classification models based on V3 sequence information. For this, 2,109 DNA sequences of V3 region from HIV-1 subtype B were used. Data labeled with geno2pheno’s results were then modeled by methods such as logistic regression, naive Bayes and random forest. All classifiers presented good predictive outputs, however random forest models showed the best discriminative performance, in the form of significant AUC results. These outcomes encourage us to continue the development of an easy to use and accurate algorithm for HIV-1 tropism diagnosis, capable of guiding clinical decision making regarding the use of coreceptor antagonists in HIV-1 treatment.