Machine Learning and Weighted Gene Coexpression Network–Based Identification of Biomarkers Predicting Immune Profiling and Drug Resistance in Lung Adenocarcinoma

Abstract

Background: The prognosis for lung adenocarcinoma (LUAD) is poor, and the recurrence rate is high. Thus, to evaluate patients’ prognoses and direct therapy choices, new prognostic markers are desperately needed.

Methods: First, gene modules associated with LUAD were identified by weighted gene coexpression network analysis (WGCNA) analysis. The expression profiles obtained were intersected with the differential expressed genes taken between LUAD samples and paracancerous samples. Afterward, stepwise regression analysis and the LASSO were used to compress the genes further, and a risk model was created. Furthermore, a nomogram based on risk scores and clinical features was created to validate the model. After that, the distinctions between the pertinent biological processes and signaling pathways among the various subgroups were investigated. Additionally, drug sensitivity testing, immunotherapy, immune infiltration analysis, and enrichment analysis were carried out. Finally, the biological role of ANLN in LUAD was explored by qPCR, cell scratch assay, and transwell.

Results: A total of 257 intersected genes were obtained by taking the intersection of the differential genes between 2866 LUAD samples and paraneoplastic samples with the module genes after we screened two particular modules that had the strongest link with LUAD by WGCNA. ANLN, CASS4, and NMUR1 were found to be distinctive genes for the development of risk models after the intersecting genes were screened to find 176 genes linked to the prognosis for LUAD. Based on risk assessments, high- and low-risk groups of LUAD patients were divided. Low-risk patients exhibited a significantly higher overall survival (OS) than those in the high-risk group. Expression of model genes correlates with infiltration of the vast majority of immune cells. Significant differences in the biological pathways, immune microenvironment, and abundance of immune cell infiltration were found between the two groups. The drug sensitivity study showed that patients in the high-risk group had higher IC₅₀ values for BMS-754807_2171 and Doramapimod_10424. Finally, in vitro experiments demonstrated that knocking down ANLN noticeably inhibited the viability, migration, and invasion of A549 cells.

Conclusion: This study may provide a theoretical reference for future exploration of potential diagnostic and prognostic biomarkers for LUAD.