Machine learning identifies MiRNA biomarkers and immune mechanisms in active tuberculosis.

Tuberculosis (TB), caused by Mycobacterium tuberculosis (Mtb), remains a major global public health threat. The rising prevalence of HIV/TB co-infection and multidrug-resistant tuberculosis (MDR-TB) has further intensified this challenge. This study aims to explore the role of microRNAs (miRNAs) in the immune response to Mtb infection and to identify potential miRNA biomarkers for active TB diagnosis using machine learning techniques. miRNA expression profiles were retrieved from the Gene Expression Omnibus (GEO) database (accession number: GSE70425). Differential expression analysis between active TB and latent TB infection (LTBI) patients was conducted using the "limma" package, with a screening threshold of |logFC| > 0.25 and p-value&#x2009;<&#x2009;0.05. Key differentially expressed miRNAs (DE-miRNAs) were further refined using machine learning algorithms, including least absolute shrinkage and selection operator (LASSO) regression, support vector machine-recursive feature elimination (SVM-RFE), and Boruta. TargetScan was employed to predict miRNA target genes, and a regulatory network was visualized using Cytoscape. Subsequently, Gene Ontology (GO) and Kyoto Encyclopedia of Genes and Genomes (KEGG) pathway enrichment analyses were conducted to elucidate the functional roles of target mRNAs. Nine machine learning models were developed based on the selected miRNAs, and their predictive performance was assessed using metrics including AUROC, accuracy, sensitivity, specificity, positive predictive value (PPV), and negative predictive value (NPV). In vitro experiments were performed using THP-1 macrophages to establish an Mtb infection model. Cells were transfected with miR-3607-3p mimics and negative controls, followed by flow cytometry for apoptosis detection, Western blot for protein expression analysis, and Quantitative Real-Time PCR (qRT-PCR) for validation of apoptosis-related gene expression. A total of 72 differentially expressed miRNAs were identified. Key miRNAs, including hsa-miR-3607-3p, hsa-miR-148b, and hsa-miR-519e, were identified through multiple machine learning methods, with hsa-miR-3607-3p emerging as the primary candidate miRNA. The nine machine learning models exhibited robust predictive performance in both the training and test sets. PPI network and functional enrichment analyses indicated that the target genes of hsa-miR-3607-3p are primarily associated with cell growth and apoptosis-related pathways. In vitro experiments further suggested that hsa-miR-3607-3p may modulate the apoptotic response of THP-1 cells to Mtb infection via a caspase-dependent mechanism. miR-3607-3p was upregulated in active TB patients and may modulate the apoptosis of THP-1 cells during Mtb infection via a caspase-dependent pathway. The mechanism of this miRNA offers preliminary insights into the immune regulation of tuberculosis. miR-3607-3p may serve as a potential biomarker for the early diagnosis and intervention of active TB; however, its clinical applicability necessitates further validation through larger sample sizes and multicenter studies.