Abstract
BACKGROUND: Uveal melanoma (UM) has favorable local tumor control, but once metastasis develops, the prognosis is rather poor. Thus, it is urgent to develop metastasis predicting markers. OBJECTIVE: Our study investigated a novel gene expression-based signature in predicting metastasis for patients with UM. METHODS: In the discovery phase, 63 patients with UM from GEO data set GSE22138 were analyzed using the Weighted Correlation Network Analysis (WGCNA) to identify metastasis-related hub genes. The Least Absolute Shrinkage and Selection Operator (Lasso) Cox regression was used to select candidate genes and build a gene expression signature. In the validation phase, the signature was validated in The Cancer Genome Atlas database. RESULTS: Forty-one genes were identified as hub genes of metastasis by WGCNA. After the Lasso Cox regression analysis, eight genes including RPL10A, EIF1B, TIPARP, RPL15, SLC25A38, PHLDA1, TFDP2, and MEGF10 were highlighted as candidate predictors. The gene expression signature for UM (UMPS) could independently predict MFS by univariate and multivariate Cox regression analysis. Incorporating UMPS increased the AUC of the traditional clinical model. In the validation cohort, UMPS performed well in predicting the MFS of UM patients. CONCLUSIONS: UMPS, an eight-gene-based signature, is useful in predicting prognosis for patients with UM.