A predictive model for evaluating the risk of latent tuberculosis relapse via machine learning

BACKGROUND: Reactivation of latent tuberculosis infection (LTBI) is a major obstacle to tuberculosis eradication. Predicting LTBI relapse is crucial for effective disease management but remains underexplored. METHODS: We analysed gene expression data from the GSE54992 dataset via differential expression analysis and weighted gene co-expression network analysis (WGCNA) to identify genes associated with LTBI relapse. A risk-predictive model was developed using three genes (ALG2, FARS2, and PGP) selected through a support vector machine with recursive feature elimination (SVM-RFE), least absolute shrinkage and selection operator (LASSO), and random forest algorithms. Model performance was assessed via the area under the receiver operating characteristic curve (AUC). Gene set enrichment analysis (GSEA) was employed to predict correlations between gene activities and immune pathways in monocytes, and gene expression was validated in infected macrophages. RESULTS: We identified 279 co-differentially expressed genes between LTBI and active tuberculosis infection (ATBI) patients. A final set of three predictive genes (ALG2, FARS2, and PGP) was selected, yielding AUC values of 0.781 and 0.725 for validation. GSEA revealed that ALG2 was associated with immune response pathways, whereas FARS2 and PGP were linked to energy and lipid metabolism in LTBI. The expression levels of these three genes were correlated with the BCG load. CONCLUSION: This integrative model provides a robust tool for assessing LTBI reactivation risk and supports more personalized tuberculosis management strategies.