Causal inference of glucocorticoid signaling in non-small cell lung cancer: integrating Mendelian randomization, single-cell transcriptomics, and imaging data

Glucocorticoid-related genes (GRGs) may facilitate tumor immune escape via tumor microenvironment (TME) modulation, but their causal association with non-small cell lung cancer (NSCLC) remains unclear. Herein, we performed Mendelian randomization analysis based on cis-eQTL data from GTEx and eQTLGen databases, combined with SMR and Bayesian colocalization analyses, to identify GRGs with significant causal links to NSCLC. TCGA transcriptomic data verified the differential expression of core GRGs. Single-cell transcriptomic analysis revealed elevated GRGs cellular module scores in M2 macrophages of NSCLC tissues relative to adjacent normal tissues, while the MuSiC algorithm estimated cell-type abundance in bulk transcriptomic datasets. We further integrated three intersecting genes (MRPS7, IQGAP1, EXOC2), M2 macrophage abundance, clinical parameters and radiomics features (extracted via nnU-Net-based segmentation) to construct a multi-omics Cox prognostic risk model. This model achieved an AUC > 0.9 for 1-year survival prediction in the training cohort and maintained robust performance (AUC 0.7–0.9) for 1–5-year survival in the independent test cohort. NSCLC tissue microarrays combined with cellular functional assays confirmed that MRPS7 not only exhibits aberrant overexpression in both NSCLC tissues and cell lines but also reveals a potential role in promoting the progression of NSCLC. This is the first study to systematically elucidate the potential causal relationship between GRGs and NSCLC risk, which may be mediated by regulating M2 macrophage function, providing novel biomarkers for precise prognosis prediction and uncovering potential mechanisms underlying NSCLC oncogenesis.