Discovery and validation of novel plasma protein biomarkers for severe tuberculosis patients.
OBJECTIVE: Severe tuberculosis (STB) imposes a substantial disease burden, yet reliable biomarkers for distinguishing STB from mild/moderate tuberculosis (MTB) remain scarce. This study aimed to identify and independently validate plasma protein biomarkers associated with tuberculosis severity. METHODS: In this multicenter prospective study, 298 adults with confirmed pulmonary tuberculosis were enrolled into screening (n = 128) and independent validation (n = 170) cohorts. Plasma samples were analysed using data-independent acquisition proteomics. Differentially expressed proteins were screened via Limma and four machine-learning algorithms, with candidate proteins measured by enzyme-linked immunosorbent assays. Receiver operating characteristic analysis assessed individual and combined diagnostic performance. RESULTS: STB patients were older and presented with lymphopenia, hypoalbuminemia, neutrophilia, and elevated lactate dehydrogenase. Among 166 differentially expressed proteins, HSPA5, HSP90B1, EEF1D, and SULT1A1 were selected for validation. In STB patients, HSPA5, HSP90B1, and EEF1D were upregulated, whereas SULT1A1 was downregulated. The four-protein panel achieved an AUC of 0.908 (95% CI 0.864-0.952), with 87.5% sensitivity and 83.8% specificity, modestly outperforming HSPA5 alone (AUC = 0.894). Functional enrichment implicated cholesterol metabolism, immune-inflammatory pathways, and endoplasmic reticulum stress. CONCLUSIONS: The four-protein panel effectively discriminated STB from MTB; however, its marginal improvement over HSPA5 alone suggests that an HSPA5-based assay may offer a simpler, more practical, and potentially cost-effective strategy for severity stratification.