Abstract
Machine learning has provided novel tools for analysis of multi-omics data for subgroups recognition in cancer to reach a clinically meaningful classification of cancer and identification of potential biomarkers. In this work, we retrieved mRNA, lncRNA, miRNA and protein expression data of non-small cell lung cancer (NSCLC) samples and used different machine learning methods for biomarker selection, diagnostic validation, construction of competing endogenous RNA network, identification of the hub axes and drug prediction. Integration of multi-omics data and machine learning resulted in identification of CDK1, TOP2A, AURKA, TPX2, BUB1B, and CENPF as key biomarkers in NSCLC. We also identified the PVT1/miR-143-3p/CDK1 axis and its associated transcription factors (FOXC1, YY1, and GATA2) as a potential regulatory network for additional investigations. These findings increase the understanding of NSCLC molecular processes and provide a foundation for developing targeted therapies and diagnostic tools.