Caracterização de biomarcadores moleculares diferenciais entre DPOC, asma e controles utilizando técnicas proteômica e FTIR.

Abstract

Introduction: Chronic Obstructive Pulmonary Disease (COPD) and asthma are chronic respiratory diseases with high morbidity, mortality, and significant socioeconomic impact worldwide, with high underdiagnosis rates. Differentiating these conditions in adults can be challenging due many factors such as late onset and smoking habit in asthma patients. Objective: To identify the metabolic profiles of patients with COPD and asthma using proteomics and infrared spectroscopy (ATR-FTIR) techniques, and to assess the ability of these methods to distinguish COPD, asthma, and control groups (smokers and non-smokers) through machine learning models. Additionally, to explore differentially expressed proteins (DEPs) in known metabolic pathways via Reactome and PANTHER. Methods: Label-free proteomics analysis was performed on serum samples from four groups: COPD, asthma, smokers, and non-smokers (10 participants each). DEPs were analyzed using unsupervised (PCA) and supervised (Logistic Regression) models, as well as molecular pathway enrichment with Reactome and PANTHER tools. ATR-FTIR was conducted on saliva samples (19 asthma, 19 COPD, 15 smokers, and 9 non-smokers). Some machine learning models were applied to classify the groups based on spectral patterns. Pairwise comparisons were performed (smokers vs. non-smokers, COPD vs. smokers, asthma vs. non-smokers, and COPD vs. asthma) to create a diagnostic flow named OLOT (Obstructive Lung Oral Test). Results: A total of 1469 proteins were identified in serum, with 29 found to be differentially expressed, 5 of which were highlighted for COPD differentiation and used in PCA and supervised Logistic Regression models. The supervised model diagnosed COPD with an accuracy above 80% in all comparisons (asthma and controls). Reactome and PANTHER analysis identified two common pathways: Wnt and integrins, associated with epithelial-mesenchymal remodeling and extracellular matrix collagen, respectively. In ATR-FTIR analysis, Neural Networks outperformed in distinguishing COPD vs. asthma and patients vs. controls (Sensitivity >90%, Specificity >70%). Logistic Regression was superior in differentiating smokers and non-smokers (Sensitivity 97%, Specificity 100%). ROC curves confirmed model efficacy (AUC >87%) across all comparisons. Conclusion: Serum proteomics and saliva ATR-FTIR, combined with machine learning, are promising diagnostic tools for COPD and asthma. Serum proteomics identified potential molecular biomarkers for COPD. Keywords: Molecular biomarkers; COPD; Asthma; Proteomics; FTIR.