Pseudotime‐Derived Seven‐Gene Prognostic Signature and TYRP1 in Melanoma
Xiaoqing Li, Yilin Deng, Yue Song, Xiaohui Liu, Hongli LiBackground
Melanoma is characterized by high malignancy and complex immune microenvironment heterogeneity. A robust prognostic framework that integrates tumor cell state diversity with clinical outcome remains lacking.
Methods
Integrative scRNA‐seq analysis of public melanoma datasets was performed using Harmony batch correction, unsupervised clustering, and Monocle 3 pseudotime trajectory reconstruction to characterize tumor cell state transitions. Pseudotime‐dynamic genes were subjected to univariate Cox screening across four independent cohorts (GSE19234, GSE22153, GSE65904, TCGA), followed by LASSO‐Cox regression to construct a multigene prognostic model. Validation encompassed immune infiltration estimation, pathway activity scoring, tumor mutational burden (TMB) analysis, and protein‐level expression assessment.
Results
scRNA‐seq identified 14 major cell populations, within which tumor cell reclustering resolved seven functional states: melanocytic, neural crest‐like, MHC‐II high, stress/transitional, proliferating, IFN‐stimulated, and invasive. Pseudotime analysis reconstructed a directional evolution trajectory originating from the melanocytic state, with the stress/transitional state acting as a central hub. Gene module enrichment further delineated the core transcriptional programs of each state. A seven‐gene prognostic model (GBP4, ITGB4, GBP2, TNFSF13B, TSC22D3, CIITA, TYRP1) achieved consistent survival stratification across all four cohorts (all p < 0.01) and outperformed 76 published signatures by C‐index ranking. High‐risk patients exhibited an immunosuppressive microenvironment, activation of protumorigenic pathways, and broad suppression of the cancer–immunity cycle. Risk score and TMB were statistically independent, yet their combined stratification significantly refined prognostic resolution ( p < 0.001). TYRP1, the sole risk gene in the model, demonstrated tumor‐specific protein upregulation by IHC and consistently predicted poor overall survival across three independent transcriptomic cohorts.
Conclusion
This study presents an integrated single‐cell and multicohort prognostic framework for melanoma, revealing the mechanistic link between tumor cell state evolution and immune microenvironment remodeling, and establishing TYRP1 as an independent prognostic biomarker to guide precision immunotherapy decision‐making.