A prognostic model for uveal melanoma in Asian populations: a comparative analysis of clinical features and gene expression patterns using the TRACE and TCGA data

Abstract

Abstract Background Uveal melanoma (UM), the most common primary intraocular malignancy in adults, shows racial disparities in incidence, genetic drivers, and clinical outcomes. While most prognostication models are based on Caucasian populations, Asians demonstrate distinct molecular profiles, necessitating population-specific risk stratification. Methods This study analyzed 53 Asian UM tumors, 17 normal choroidal tissues (TRACE database), and 80 Caucasian UM samples (TCGA database). Differential gene analysis, immune microenvironment profiling, and survival modeling were performed. A 7-gene prognostic signature was developed by LASSO regression and validated across cohorts, with drug sensitivity predicted using GDSC2 data. Results The Asian UM exhibited 3,827 tumor-specific differentially expressed genes (DEGs) compared to the normal choroid, with enrichment in PI3K-Akt signalling, and 3,814 race-specific DEGs compared to Caucasians, suggesting specific disease pathways and variations in the tumor microenvironment. The tumor microenvironment in Asian UM exhibited increased immunological activation (M1 macrophages, PD-L1, CD8; p < 0.05), while Caucasian uveal melanoma was marked by immunosuppressive predominance (M2 macrophages, MDSCs). The seven-gene prognostic model (MMP2, LRAT, NOG, IHH, CDH18, MYH11, and SELE) exhibited strong predictive efficacy in Asians (AUC: 0.979, 0.924, and 0.984 for 1, 3, and 5-year survival) but was less successful in Caucasians. High-risk scores correlated with metastasis (12/26 vs. 4/27; p = 0.02) and had independent prognostic value. Conclusions This cross-racial UM study reveals significant molecular and immune differences, indicating that Asian UM may be more responsive to immunotherapy The population-specific prognostic model improves our understanding of molecular differences in Asian and Caucasian UM, warranting further validation in multiethnic cohorts.