Driver Mutation Subtypes Differentially Shape Immune Evasion Landscapes in Melanoma: An AI-Driven Inflammatory Pathway Model Implicating CCNE1.

Background: Melanoma harbors highly heterogeneous tumor immune microenvironments shaped by driver mutations in BRAF, NRAS, and NF1. How mutational context modulates inflammatory signaling and immune evasion mechanisms of prognosis-related genes remains poorly understood. Methods: ssGSEA scored 15 inflammatory pathways across four cohorts (TCGA, GSE19234, GSE22153, and GSE65904). Cross-cohort univariate Cox regression and a 10-algorithm machine learning framework identified and optimized a prognostic model. Immune microenvironment differences across BRAF, NRAS, NF1, and Triple-WT subtypes were characterized using ESTIMATE, CIBERSORT, and GSVA. CCNE1 was validated by shRNA knockdown in melanoma cell lines, and a mutation subtype-specific virtual knockdown model was constructed from scRNA-seq data. Results: The prognostic model achieved a 5-year AUC of 0.95 in TCGA and outperformed published signatures in three of four cohorts. High-risk patients showed markedly reduced immune infiltration (ImmuneScore Conclusion: This study establishes a high-performance inflammatory pathway-based prognostic model for melanoma and demonstrates that driver mutation subtypes differentially shape the immune microenvironment landscape. CCNE1 functions as a key oncogenic immune regulator whose modulation of antigen presentation and immune checkpoint expression is contingent on mutational context, most prominently in NF1-mutant tumors, underscoring the value of mutation-informed personalized immunotherapy strategies in melanoma.