Biofluidos e espectrometria de massas para triagem de pacientes para COVID-19

• Ano de defesa: 2024
• Autor(a) principal: Almeida, Camila Medeiros de
• Tipo de documento: Tese
• Tipo de acesso: Acesso aberto
• Idioma: por
• Instituição de defesa: Universidade Federal do Espírito Santo; BR; Doutorado em Química; Centro de Ciências Exatas; UFES; Programa de Pós-Graduação em Química
• Palavras-chave em Português: Biofluidos; Química analítica; Espectrometria de massa; Quimiometria; Aprendizagem de máquina; COVID-19; Biofluids; Mass spectrometry; Machine learning; Química
• Link de acesso: http://repositorio.ufes.br/handle/10/17639

The COVID-19 disease has been and continues to be a global health concern. The identification of infected patients through rapid and efficient screenings remains necessary to contain its spread. Biological fluids, such as serum and saliva, offer ease of collection and provide rich information about molecular changes in the body during illness. The use of mass spectrometry (MS) combined with machine learning (ML) has been applied to biofluids from patients with diseases and controls to identify biomarkers and conduct rapid and effective screenings. Therefore, this thesis aims to present advancements in the search for disease biomarkers, particularly for COVID-19, using technologies based on Matrix-Assisted Laser Desorption Ionization Mass Spectrometry (MALDI MS) and Electrospray Ionization Mass Spectrometry (ESI MS), along with chemometric data treatments. To achieve this, a methodology was developed for screening patients suspected of having COVID-19 based on saliva samples, using MALDI MS with the assistance of Support Vector Machine (SVM) learning. This involved optimizing sample preparation and analysis parameters. The most efficient results in a shorter analysis time were obtained by digesting saliva with 10 μL of trypsin for 2 hours. Optimization of the parameters at 1M resolution was ideal for the analyses. SVM models were created using data from the analysis of 149 samples, 97 positive and 52 negative for COVID-19. Two models yielded the best results. SVM1 selected 780 variables with a false negative rate (FNR) of 0%, while SVM2 selected only 2 variables (525.4 Da and 1410.8 Da) with a 3% FNR. Another application of MS in biofluids was the development of a multiomic method for screening patients infected with SARS-CoV-2 based on serum lipid and proteomic profiles. ESI MS was used to investigate the lipid profile of 239 serum samples (119 positive and 120 negative for COVID-19). MALDI MS was used to analyze the proteomic profile of 300 serum samples (150 positive and 150 negative for COVID-19). After processing MS data and variable selection, statistical analyses such as Volcano plot, Heatmap, principal component analysis (PCA), partial least squares discriminant analysis (PLS-DA), and SVM were performed to distinguish the most relevant variables for classifying positive and negative samples for COVID-19. In lipidomic analyses using ESI(±)-Orbitrap MS and SVM models, sensitivities of 96.67% and 100%, specificities of 82.14% and 96.88%, and accuracies of 89.66% and 98.44% were observed for positive and negative ion mode analyses, respectively. In proteomic analyses using MALDI(+) MS, the linear PLS-DA model demonstrated an accuracy of 99.10%. Thus, the combination of MS techniques with chemometric data treatments has shown promising alternatives with high sensitivity and specificity to discriminate infected and non-infected biological samples by SARS-CoV-2