Investigação soro-epidemiológica de fatores preditivos associados ao perfil de resposta imune humoral da COVID-19 em doadores de sangue de Fortaleza

• Ano de defesa: 2024
• Autor(a) principal: Aksenen, Cleber Furtado
• Tipo de documento: Dissertação
• Tipo de acesso: Acesso aberto
• Idioma: por
• Palavras-chave em Português: Teste Sorológico para COVID-19; Vigilância em Saúde Pública; Testes Sorológicos; Aprendizado de Máquina Supervisionado; COVID-19 Serological Testing; Public Health Surveillance; Serologic Tests; Supervised Machine Learning; CNPQ::CIENCIAS BIOLOGICAS
• Link de acesso: http://repositorio.ufc.br/handle/riufc/78424

The evolution of Severe Acute Respiratory Syndrome Coronavirus 2 (SARS-CoV-2) has led to the emergence of increasingly adapted variants to the epidemiological landscape, in addition to fluctuations in the number of cases and deaths from the disease (COVID19) during critical moments of the pandemic. Multiple vaccination campaigns with different combinations of technologies, along with successive reinfections/exposures to the virus, have resulted in variations in individual seroepidemiological profiles, highlighting the multivariate nature of the immune system's defense mechanisms. Therefore, the investigation of specific anti-SARS-CoV-2 antibodies becomes relevant for assessing the level of humoral response and seroconversion resulting from successive exposures and cycles of population immunization. In this context, the study was conducted with 17,904 donors between 2020 and 2024 in Fortaleza, along with active recruitment of a subgroup of participants. A comparison of sociodemographic, temporal, and immunological data, including self-reported responses, hematological, biochemical, and serological tests for a prospective subgroup, was performed. Independent statistical analyses of the variables were conducted using Mann-Whitney tests for two groups and Kruskal-Wallis tests for more than two groups, along with post-hoc analyses with Bonferroni corrections and effect size calculations. A multivariate investigation was included, using three supervised learning models: Support Vector Machine (SVM), Logistic Regression (LR), and Gradient Boosting (GB), to predict the stratification of individual profiles into different outcomes. The evaluated participants were representative of the 12 regions of Fortaleza, with broader coverage of the area near the blood center. The majority were aged between 16 and 29 years, predominantly male, of mixed race, and with completed high school education. There was a statistical association between antibody titers when compared across different periods and vaccine doses, as well as multiple correlations between sociodemographic variables. The GB model showed the best performance in predicting outcomes for the dataset (2020 to 2024). Vaccination, primarily with the Pfizer vaccine, stood out as a key feature for the models. This work emphasizes the importance of using a high-granularity dataset to select candidates, aiming for a better understanding of the seroepidemiological variability in relation to COVID-19.