Método de aprendizado de máquina não supervisionado aplicado ao estudo da propagação de ciclo único do HIV em cultivo celular

• Ano de defesa: 2019
• Autor(a) principal: Fabreti, Luiza Guimaraes [UNIFESP]
• Tipo de documento: Dissertação
• Tipo de acesso: Acesso aberto
• Idioma: por
• Instituição de defesa: Universidade Federal de São Paulo (UNIFESP)
• Palavras-chave em Português: HIV; Infecções Por HIV; Análise Estatística; Filogenia
• Link de acesso: https://sucupira.capes.gov.br/sucupira/public/consultas/coleta/trabalhoConclusao/viewTrabalhoConclusao.jsf?popup=true&id_trabalho=8134321; https://repositorio.unifesp.br/handle/11600/59208

The human immunodeficiency vírus (HIV) has a high genetic diversity due mostly to the lack of error repair mechanism of the viral reverse transcriptase combined with the quick replicative dynamics of the virus. These factors associated to the selective pressure of the environment contribute to a rapid evolution of the viral population within the host. To enter the cell, HIV uses both a cellular receptor (CD4) and a chemokine coreceptor. The most important coreceptors for the infection are CCR5 and CXCR4, the former being associated with the onset of infection and the second with late phase of infection. In the present study we have applied an unsupervised machine learning method known as principal component analysis (PCA) and the phylogenetic inference to estimate the genetic diversification of the population after a single replication cycle. To do this, the data from statistical inference have been used to estimate the probability of each nucleotide for each genome position of HIV, in different experimental conditions. PCA compared mutation patterns when infection occurs on CD4 T lymphocytes and peripheral blood mononuclear cells (PBMCs) in different (stimulated and non-stimulated) cell states and for the two viral types that utilize different coreceptors. In all cases, both PCA and phylogenetic tree produced clusters accordingly. This shows that after a single cycle of replication the populations presented differences of variability. It has also been observed that the genome positions that mutate are equally distributed throughout the genome, with an average range of 20 positions in the population analysis.