Circadian-immune-related gene signature for lung squamous cell carcinoma: machine learning and multi-omics analysis

Circadian disruption promotes tumor progression and therapy resistance, but its prognostic value and impact on the tumor immune microenvironment in lung squamous cell carcinoma (LUSC) remain unclear. This study aimed to develop a circadian-immune-related gene signature to improve LUSC risk stratification and therapy guidance.

We integrated multi-omics data from LUSC patients (n=494) across The Cancer Genome Atlas and Gene Expression Omnibus (GEO) database GSE73403 (n=69). A circadian-immune-related gene prognostic signature (CIGPS) was constructed from 1,677 circadian rhythm-related genes through weighted gene co-expression network analysis (WGCNA) and ten machine learning algorithms (StepCox + GBM optimized) and externally validated. Bioinformatics analyses assessed the tumor immune microenvironment, mutation landscape, and therapy response. Single-cell RNA sequencing data (GSE148071) explored gene expression at cellular resolution.

A six-gene CIGPS was developed, comprising CLDN1, FGG, ALPL, TREM1, NKX2-1, and CLDN5. This signature stratified patients into high- and low-risk groups with divergent overall survival in both cohorts. Unsupervised clustering revealed C1/C2 subtypes: C1 displayed an immune-hot yet excluded phenotype with higher CD8+ T cell infiltration but elevated physical barriers; C2 exhibited an immune-cold phenotype with DNA replication and epigenetic machinery activation. Both subtypes demonstrated distinct therapeutic vulnerabilities and divergent drug sensitivity profiles. Single-cell profiling demonstrated heterogeneous spatial distribution of signature genes across distinct tumor microenvironmental compartments.

The CIGPS serves as a potent prognostic tool that elucidates the heterogeneous tumor immune microenvironment in LUSC. It holds significant promise for guiding risk assessment and informing personalized therapeutic strategies based on distinct molecular subtypes.