Mitochondrial non-coding RNAs as novel biomarkers and therapeutic targets in lung cancer integration of traditional bioinformatics and machine learning approaches.

BACKGROUND

Lung cancer diagnosis requires cost-effective biomarkers. Mitochondrial non-coding RNAs (mtRNAs) represent unexplored diagnostic targets.

METHODS

We analyzed TCGA-LUAD/LUSC miRNA-seq data to identify mtRNAs via mitochondrial genome alignment. Machine learning algorithms (SVM, Random Forest, Logistic Regression) classified samples using differentially expressed mtRNAs (P < 0.01, |log2FC| > 1). Top-ranked t00043332 was functionally validated in A549/PC9 cells.

RESULTS

Ten mtRNAs distinguished cancer from normal tissues. Random Forest and Logistic Regression achieved superior classification (AUC > 0.92) versus SVM. Nine mtRNAs were upregulated, one downregulated in cancer. No survival associations were observed. t00043332 overexpression promoted proliferation, migration, invasion, and apoptosis resistance.

CONCLUSION

mtRNAs serve as effective lung cancer diagnostic biomarkers through integrated traditional and AI approaches. t00043332 functions as an oncogene, providing therapeutic targets and advancing biomarker discovery.