Abstract:

Objective: To investigate the correlation between key genes associated with ferroptosis and the pathogenic mechanism of pulmonary tuberculosis using transcriptomics and machine learning methodologies. Methods: Two transcriptomic datasets were obtained, named as GSE153326 and GSE67589, through searching the NCBI GEO public repository (http://www.ncbi.nlm.nih.gov/geo) using keywords “pulmonary tuberculosis” and specific criteria such as sequencing type (transcriptomics) and species (HOMO sapiens). GSE153326 served as a training dataset, comprising eight blood samples from healthy individuals and 52 blood samples with positive Mycobacterium tuberculosis (MTB). GSE67589 served as the validation dataset, including 30 blood samples from healthy individuals and 27 MTB-positive samples. Data refinement and annotation were performed using R scripts. After identification of differentially expressed genes in the two transcriptomic datasets, gene expression related to ferroptosis (TBFerDEG) in the training dataset GSE153326 were obtained. Enrichment analysis of Gene Ontology (GO) and Kyoto Encyclopedia of Genes and Genomes (KEGG) pathways for TBFerDEG was conducted. LASSO regression analysis and the SVM algorithm were applied to extract key genes associated with ferroptosis from TBFerDEG. ROC analysis was performed to explore the drug regulatory network related to these key genes. Finally, the key genes were incorporated into the validation dataset GSE67589 to validate the diagnostic performance of the ferroptosis-associated key genes identified in the training dataset. Results: Through bioinformatics analysis, a total of 416 TBFerDEGs were identified and 56 differentially expressed genes were obtained after filtering of non-matched genes. GO enrichment analysis revealed that biological processes related to ferroptosis in pulmonary tuberculosis include cellular response to chemical stress by autophagy regulation, mitochondrial autophagy, and mitochondrial disintegration. The implicated pathways encompassed the AMPK signaling pathway and ferroptosis. Through LASSO regression analysis and the SVM algorithm, five key genes associated with ferroptosis were ultimately identified, BID, AR, STK11, ALOX12, and SRC, with AUCs of 0.807, 0.858, 0.734, 0.840, and 0.880, respectively. Expression of AR (P=0.004) and SRC (P=0.017) in MTB-positive group were significantly different compared to the control group in the validation dataset. Conclusion: AR and SRC are key genes associated with ferroptosis in pulmonary tuberculosis, providing valuable insights for future basic research in this field.

Key words: Tuberculosis, pulmonary, Ferroptosis, Gene expression, Machine learning