Aprendizado de máquina aplicado à vigilância genômica de vírus emergentes e reemergentes

• Ano de defesa: 2022
• Autor(a) principal: Álvaro Salgado de Abreu
• Tipo de documento: Tese
• Tipo de acesso: Acesso aberto
• Idioma: por
• Instituição de defesa: Universidade Federal de Minas Gerais; Brasil; ICB - INSTITUTO DE CIÊNCIAS BIOLOGICAS; Programa de Pós-Graduação em Bioinformatica; UFMG
• Palavras-chave em Português: aprendizado de máquina; vigilância genômica; Bioinformática; Aprendizado de Máquina; Genômica; Inteligência Artificial; Vírus da Febre Amarela; Vírus da Dengue
• Link de acesso: http://hdl.handle.net/1843/55401

Arboviruses transmitted by mosquitoes present major challenges for public health and are responsible for epidemics which cause significant impacts on the health system of the countries where they occur, with a diverse spectrum of clinical conditions. Among them, yellow fever and dengue viruses are of special interest in the Brazilian territory. Yellow fever virus is responsible for the most severe disease transmitted by mosquitoes in the tropics, and Brazil has faced recent outbreaks with high mortality rates in areas where the virus had not been reported for decades. These sites consist of densely populated urban regions with a high prevalence of unvaccinated people. Dengue virus, in turn, is a threat that puts a third of the global population at risk, especially in places where its main vector, mosquitoes of the Aedes genus, is more prevalent. In Brazil, specifically, different serotypes have caused major outbreaks in recent decades, so that genomic surveillance comes up as an important measure for early detection of emerging and reemerging viruses, as well as for investigating their dynamic behavior and dissemination. In this context, data analysis tools based on machine learning are able to extract useful information from large volumes of data and assist in the objectives of genomic surveillance. In the work presented here, we analyzed complete and almost complete genomic sequences of yellow fever and dengue viruses, associated with clinical, laboratory, epidemiological, geographic and temporal data, in order to identify genetic signatures correlated with observed biological characteristics. As a result, we identified non-synonymous nucleotide variations associated with the cycle threshold of yellow fever samples from non-human primates and the severity of yellow fever infections in humans, for which we performed in-silico structural protein modeling and discussed possible biological implications. In addition, we also identified genetic signatures that differentiate strains of dengue virus serotype 2 in a recent outbreak in Brazil, highlighting the method’s complementarity and accordance to phylogenetic analysis. Therefore, this work presents an initial, versatile and fast approach to assist in real-time genomic surveillance.