Discovering Biomarkers for Asymptomatic Tuberculosis via Olink Proteomics and Machine Learning

The diagnosis of asymptomatic tuberculosis (TB) remains challenging due to an early disease stage. This study aimed to identify and validate plasma biomarkers for asymptomatic TB by integrating the Olink proteomics with multiple machine learning algorithms. Plasma samples were analyzed using the Olink Proximity Extension Assay targeting 92 inflammation-related proteins; support vector machine (SVM), random forest, neural network, and XGBoost algorithms were employed to screen and identify the most discriminative biomarkers. Our data revealed that EN-RAGE and MCP-3 were significantly upregulated in asymptomatic TB cases. The combination of EN-RAGE and MCP-3 could accurately discriminate asymptomatic TB from healthy controls and latent TB infection (LTBI), yielding an area under the curve (AUC) of 0.90 (95% CI: 0.85–0.95). ELISA validation performed in an independent cohort confirmed significant elevations of EN-RAGE and MCP-3 in asymptomatic TB compared to healthy controls and LTBI (AUC = 0.837, 95% CI: 0.75–0.924, p < 0.05). These findings indicate that the combination of EN-RAGE and MCP-3 possesses a high potential for diagnosis of asymptomatic TB. Further translation of EN-RAGE and MCP-3 into clinical practice may facilitate early identification of asymptomatic TB, improving patient outcomes and enhancing TB control strategies.