Table_1_PI3K/AKT/mTOR pathway-derived risk score exhibits correlation with immune infiltration in uveal melanoma patients.docx

Uveal melanoma (UVM) is a rare but highly aggressive intraocular tumor with a poor prognosis and limited therapeutic options. Recent studies have implicated the PI3K/AKT/mTOR pathway in the pathogenesis and progression of UVM. Here, we aimed to explore the potential mechanism of PI3K/AKT/mTOR pathway-related genes (PRGs) in UVM and develop a novel prognostic-related risk model. Using unsupervised clustering on 14 PRGs profiles, we identified three distinct subtypes with varying immune characteristics. Subtype A demonstrated the worst overall survival and showed higher expression of human leukocyte antigen, immune checkpoints, and immune cell infiltration. Further enrichment analysis revealed that subtype A mainly functioned in inflammatory response, apoptosis, angiogenesis, and the PI3K/AKT/mTOR signaling pathway. Differential analysis between different subtypes identified 56 differentially expressed genes (DEGs), with the major enrichment pathway of these DEGs associated with PI3K/AKT/mTOR. Based on these DEGs, we developed a consensus machine learning-derived signature (RSF model) that exhibited the best power for predicting prognosis among 76 algorithm combinations. The novel signature demonstrated excellent robustness and predictive ability for the overall survival of patients. Moreover, we observed that patients classified by risk scores had distinguishable immune status and mutation. In conclusion, our study identified a consensus machine learning-derived signature as a potential biomarker for prognostic prediction in UVM patients. Our findings suggest that this signature is correlated with tumor immune infiltration and may serve as a valuable tool for personalized therapy in the clinical setting.

葡萄膜黑色素瘤（Uveal Melanoma, UVM）是一种罕见但恶性程度极高的眼内肿瘤，其预后较差且治疗手段有限。近年来的研究表明，PI3K/AKT/mTOR信号通路（PI3K/AKT/mTOR pathway）参与了UVM的发病与进展过程。本研究旨在探讨PI3K/AKT/mTOR通路相关基因（PI3K/AKT/mTOR pathway-related genes, PRGs）在UVM中的潜在作用机制，并构建一种新型的预后相关风险模型。我们通过对14份PRGs表达谱开展无监督聚类（unsupervised clustering）分析，成功鉴定出3种具有差异化免疫特征的独特亚型。其中亚型A的总生存期（overall survival）最差，且人类白细胞抗原（human leukocyte antigen, HLA）、免疫检查点（immune checkpoints）的表达水平更高，免疫细胞浸润（immune cell infiltration）程度更强。后续富集分析（enrichment analysis）结果显示，亚型A主要参与炎症反应（inflammatory response）、细胞凋亡（apoptosis）、血管生成（angiogenesis）以及PI3K/AKT/mTOR信号通路的调控。对不同亚型开展差异表达分析后，共筛选出56个差异表达基因（differentially expressed genes, DEGs），这些DEGs的主要富集通路与PI3K/AKT/mTOR信号通路密切相关。基于上述DEGs，我们构建了一种共识机器学习衍生特征标签（RSF模型, Random Survival Forest model），在76种算法组合中，该模型展现出最优的预后预测效能。该新型特征标签对患者的总生存期表现出优异的稳健性与预测能力。此外，我们发现根据风险评分进行分组的患者，其免疫状态与突变特征存在显著差异。综上，本研究鉴定出一种共识机器学习衍生特征标签，可作为UVM患者预后预测的潜在生物标志物。研究结果表明，该特征标签与肿瘤免疫浸润状态相关，有望成为临床场景下个性化治疗的有效工具。