目标：上皮间质转化（EMT）与口腔鳞状细胞癌（OSCC）不良预后有关。本研究旨在针对OSCC建立一个EMT基因签名以预测预后。方法：在TCGA数据集中，通过筛选p值< 0.05的与预后相关的EMT基因。然后，采用LASSO方法构建EMT基因签名。利用GSE41613数据集外部验证该签名在预测预后方面的有效性。通过ESTIMATE和CIBERSORT算法评估该签名与基质/免疫评分和免疫细胞浸润的相关性。利用GSEA分析高风险和低风险表型中激活的显著信号通路。通过RT-qPCR在40对OSCC和正常组织中验证每个基因的表达。结果：在OSCC中构建了一个预后性的9个EMT基因签名。高风险评分预示着较差的临床预后与低风险评分相比。ROC曲线证实了在预测1年、3年和5年生存方面的良好表现。多变量Cox分析显示该签名是独立预测OSCC预后的。该签名的良好预测效果在GSE41613数据集中得到了验证。此外，该签名与OSCC的基质/免疫评分和免疫细胞浸润密切相关。两个亚组激活了不同的信号通路。在验证后，与正常组织相比，OSCC中AREG、COL5A3、DKK1、GAS1、GPX7和PLOD2明显上调，而SFRP1下调。结论：我们的数据在OSCC中鉴定并验证了一个有力的EMT基因签名，为预测预后和针对OSCC治疗提供了一种新的临床工具。Copyright © 2023 Ai, Tan, Liu, Song, Wang, Xia and Fu.

Objective: Epithelial-mesenchymal transition (EMT) is linked to an unfavorable prognosis in oral squamous cell carcinoma (OSCC). Here, we aimed to develop an EMT gene signature for OSCC prognosis. Methods: In TCGA dataset, prognosis-related EMT genes with p < 0.05 were screened in OSCC. An EMT gene signature was then conducted with LASSO method. The efficacy of this signature in predicting prognosis was externally verified in the GSE41613 dataset. Correlations between this signature and stromal/immune scores and immune cell infiltration were assessed by ESTIMATE and CIBERSORT algorithms. GSEA was applied for exploring significant signaling pathways activated in high- and low-risk phenotypes. Expression of each gene was validated in 40 paired OSCC and normal tissues via RT-qPCR. Results: A prognostic 9-EMT gene signature was constructed in OSCC. High risk score predicted poorer clinical outcomes than low risk score. ROCs confirmed the well performance on predicting 1-, 3- and 5-year survival. Multivariate cox analysis revealed that this signature was independently predictive of OSCC prognosis. The well predictive efficacy was validated in the GSE41613 dataset. Furthermore, this signature was distinctly related to stromal/immune scores and immune cell infiltration in OSCC. Distinct pathways were activated in two subgroups. After validation, AREG, COL5A3, DKK1, GAS1, GPX7 and PLOD2 were distinctly upregulated and SFRP1 was downregulated in OSCC than normal tissues. Conclusion: Our data identified and verified a robust EMT gene signature in OSCC, which provided a novel clinical tool for predicting prognosis and several targets against OSCC therapy.Copyright © 2023 Ai, Tan, Liu, Song, Wang, Xia and Fu.