关键词: 葡萄膜黑色素瘤, 肿瘤转移, 生物信息学, Lasso-Cox回归, 疾病特异生存

Abstract: Objective To identify the biomarkers of metastatic uveal melanoma (UM) and develop a predictive model for the risk of metastasis in UM patients. Design Case control study and diagnosis test. Participants The transcriptome data of UM samples were obtained from The Cancer Genome Atlas (TCGA) and Gene Expression Omnibus (GEO) database. Methods To analyze the differentially expressed genes in metastatic UM samples based on the transcriptome data UM samples obtained from the public database. To construct and validate a gene signature for predicting the risk of metastasis according to the differentially expressed genes. The differential expressed genes in metastatic UM tumor tissues were analyzed based on GSE21138 and GSE44299 data sets. The samples in GSE21138 were used as the training set. Lasso-Cox regression analysis was employed to construct a gene signature for predicting metastasis-free survival in UM patients. The receiver operating characteristic curves for predicting 1-, 3- and 5-year metastasis-free survival were generated. The Kaplan-Meier analysis was used to explore the predictive value of prognosis in patients at different risk group. Finally, the predictive value of gene signature is also validated in GSE44299 data set and TCGA database. Main Outcome Measures The differentially expressed genes between metastatic and non-metastatic UM tissues. The predictive value of the constructed gene signature for predicting metastasis-free survival, metastasis related death, overall survival and disease-specific survival in UM patients. Results A total of 17 differentially expressed genes (Fold Change＞2) were identified in metastatic UM samples. A 6-gene signature was developed by Lasso-Cox regression analysis for predicting metastasis-free survival including (RIMS2, PTP4A3, HTR2B, HNMT, COBLL1, ID2). The AUC values of 1-, 3- and 5-year metastasis-free survival predicted by the 6-gene signature were 0.78, 0.89 and 0.85, respectively. The results of Kaplan-Meier survival analysis showed that patients in high-risk group based on 6-gene signature risk score had unfavorable metastasis-free survival. The 6-gene signature also showed good predictive value for overall survival and disease-specific survival in the GSE44299 data set and TCGA database. Conclusion A 6-gene signature model based on UM metastasis related genes was developed and validated. The 6-gene signature has good predictive value for metastasis-free survival, overall survival and disease-specific survival in UM patients. (Ophthalmol CHN, 2021, 30: 412-420)