Abstract / Summary
Abstract Background : Genetic susceptibility to cervical cancer is closely associated with dysregulation of the tumor microenvironment. However, genetic mapping, causal inference, and single-cell functional characterization have rarely been integrated within a unified analytical framework, limiting the translational potential of candidate therapeutic targets. Methods : This study integrated conditional and joint association analysis (COJO), Multi-marker Analysis of GenoMic Annotation (MAGMA), two-sample Mendelian randomization (MR), differential expression analysis, chromosomal mapping, gene correlation analysis, immune infiltration analysis, gene-immune cell association analysis, machine learning, and single-cell RNA sequencing (scRNA-seq). COJO and MAGMA were first used to identify cervical cancer susceptibility genes and independent genetic signals from genome-wide association study (GWAS) data. MR was then performed to assess potential causal associations between gene expression and cervical cancer. Differential expression and genomic localization analyses were used to prioritize core genes. Their potential immune-related mechanisms were investigated through immune infiltration and gene-immune cell association analyses. Machine-learning models were used to evaluate predictive performance and construct a nomogram. Finally, scRNA-seq was used to characterize the cell-type-specific expression of the prioritized genes. Results : COJO and MAGMA identified 12 independent genetic signals corresponding to eight susceptibility genes. MR supported associations of genetically predicted expression of SERPINB13 (IVW: OR = 1.72, 95% CI = 1.48-1.99, P = 2.3 x 10^-8), PDCD1LG2 (IVW: OR = 1.58, 95% CI = 1.35-1.85, P = 7.6 x 10^-7), SERPINF2 (IVW: OR = 1.45, 95% CI = 1.23-1.71, P = 3.1 x 10^-5), and MMP1 (IVW: OR = 1.36, 95% CI = 1.15-1.61, P = 4.8 x 10^-4) with cervical cancer. All four genes were significantly upregulated in cervical cancer tissues (FDR < 0.001). Immune infiltration analysis indicated an immunosuppressive microenvironment characterized by enrichment of regulatory T cells and M2 macrophages and reduced cytotoxic immune-cell activity. The core genes were positively correlated with immunosuppressive cell populations (r > 0.5, P < 0.01). Among the evaluated machine-learning algorithms, the random forest (RF) model showed the best performance (AUC = 0.96, 95% CI = 0.93-0.98). The resulting nomogram showed good calibration (Hosmer-Lemeshow test, P = 0.78), and decision curve analysis suggested a higher net benefit than the comparator strategies across threshold probabilities of 0.1-0.8. scRNA-seq localized SERPINB13 predominantly to M2 macrophages and epithelial cells, PDCD1LG2 to T cells and macrophages, and MMP1 to epithelial and stromal cells. Conclusions : This multidimensional analysis identified SERPINB13, PDCD1LG2, SERPINF2, and MMP1 as candidate therapeutic targets and potential predictors of cervical cancer. The findings suggest that these genes may contribute to cervical carcinogenesis through immune-microenvironment regulation and may support future risk-stratification research. Experimental and prospective clinical validation is required before clinical application.