[Comprehensive bioinformatics analysis and machine learning to investigate the regulatory mechanism and clinical value of secreted phosphoprotein 1 gene in lung squamous cell carcinoma].

This study aimed to comprehensively investigate the regulatory mechanism and clinical value of secreted phosphoprotein 1 (SPP1) in the three-stage progression of "smoking-chronic obstructive pulmonary disease (COPD)-carcinogenesis" in lung squamous cell carcinoma (LUSC) through integrated bioinformatics analysis and machine learning. The datasets for the three stages of LUSC were downloaded from the Gene Expression Omnibus (GEO) database, including GSE18385 (containing lung tissue samples from 31 healthy smokers and 21 healthy non-smokers), GSE38974 (containing lung tissue samples from 23 smoking COPD patients and 9 healthy smokers), and GSE12472 (containing lung tissue samples from 18 LUSC patients with COPD and 17 smoking COPD patients). The Cancer Genome Atlas (TCGA)-LUSC dataset (comprising 504 samples, including lung tissue samples from LUSC patients and their matched normal lung tissue samples) was downloaded from TCGA database for further analysis. Samples and follow-up information from 208 non-small cell lung cancer patients who underwent radical resection and mediastinal lymph node dissection at Zhongshan Hospital, Fudan University in 2005 were selected for prognostic analysis and validation. Weighted Gene Co-expression Network Analysis (WGCNA) was applied to screen stage-specific module genes. Differentially expressed genes (DEGs) were identified through differential expression analysis. The CIBERSORT algorithm and Gene Set Enrichment Analysis (GSEA) were used to characterize the immune microenvironment. Eight machine learning algorithms and protein-protein interaction (PPI) network analysis were combined to screen for core regulatory targets. Results from WGCNA and differential analysis of the GEO datasets indicated that SPP1 is consistently highly expressed across the three stages of LUSC. Analysis of the TCGA-LUSC dataset further verified that the relative expression level of SPP1 in lung tissues of LUSC patients was significantly higher than in normal lung tissues (9.13±2.01 vs 4.68±1.64, <0.001). Furthermore, SPP1 expression was significantly higher in patients with TNM stage Ⅲ than in those with stage Ⅱ (9.59±2.09 vs 8.80±2.15, =0.045). Male LUSC patients with high smoking exposure exhibited higher SPP1 expression levels than those with low smoking exposure (9.56±2.23 vs 8.60±2.04, =0.032). Survival prognosis analysis revealed that among male patients, the difference in median overall survival (OS) between the high SPP1 expression group and the low expression group was statistically significant [2.90 (95%: 2.11-4.64) years vs 4.69 (95%: 2.95-7.34) years, =0.032). Data validation from Zhongshan Hospital, Fudan University, also showed that the 5-year survival rate of lung cancer patients with high SPP1 expression was lower than that of patients with low SPP1 expression (49.3% vs 62.6%, =0.042). Results from the CIBERSORT algorithm indicated that high SPP1 expression drives increased infiltration of M2 macrophages (<0.001). Machine learning combined with PPI network analysis identified NTN1 and CX3CL1 as key regulatory targets of SPP1, which may be associated with the occurrence and development of lung cancer. SPP1 may promote LUSC progression by mediating M2 macrophage polarization through suppressing NTN1 or activating CX3CL1, suggesting its potential as a prognostic biomarker and therapeutic target for high-risk male smokers.