Oral cancer (OC) is the most common form of head and neck cancer. Despite the high incidence and unfavourable patient outcomes, currently, there are no biomarkers for the early detection of OC. This study aims to discover, develop, and validate a novel saliva-based microRNA signature for early diagnosis and prediction of OC risk in oral potentially malignant disorders (OPMD). The Cancer Genome Atlas (TCGA) miRNA sequencing data and small RNA sequencing data of saliva samples were used to discover differentially expressed miRNAs. Identified miRNAs were validated in saliva samples of OC (n = 50), OPMD (n = 52), and controls (n = 60) using quantitative real-time PCR. Eight differentially expressed miRNAs (miR-7-5p, miR-10b-5p, miR-182-5p, miR-215-5p, miR-431-5p, miR-486-3p, miR-3614-5p, and miR-4707-3p) were identified in the discovery phase and were validated. The efficiency of our eight-miRNA signature to discriminate OC and controls was: area under curve (AUC): 0.954, sensitivity: 86%, specificity: 90%, positive predictive value (PPV): 87.8% and negative predictive value (NPV): 88.5% whereas between OC and OPMD was: AUC: 0.911, sensitivity: 90%, specificity: 82.7%, PPV: 74.2% and NPV: 89.6%. We have developed a risk probability score to predict the presence or risk of OC in OPMD patients. We established a salivary miRNA signature that can aid in diagnosing and predicting OC, revolutionising the management of patients with OPMD. Together, our results shed new light on the management of OC by salivary miRNAs to the clinical utility of using miRNAs derived from saliva samples.

口腔癌（OC）是头颈癌最常见的形式。尽管 OC 发病率高且患者预后不佳，但目前尚无用于早期检测 OC 的生物标志物。本研究旨在发现、开发和验证一种基于唾液的新型 microRNA 特征，用于早期诊断和预测口腔潜在恶性疾病 (OPMD) 的 OC 风险。利用癌症基因组图谱（TCGA）的miRNA测序数据和唾液样本的小RNA测序数据来发现差异表达的miRNA。使用定量实时 PCR 在 OC ( n = 50)、OPMD ( n = 52) 和对照 ( n = 60) 的唾液样本中验证了已识别的 miRNA。八种差异表达的 miRNA（miR-7-5p、miR-10b-5p、miR-182-5p、miR-215-5p、miR-431-5p、miR-486-3p、miR-3614-5p 和 miR- 4707-3p）在发现阶段被识别并得到验证。我们的八个 miRNA 特征区分 OC 和对照的效率为：曲线下面积 (AUC)：0.954，敏感性：86%，特异性：90%，阳性预测值 (PPV)：87.8% 和阴性预测值 (NPV) ：88.5%，而 OC 和 OPMD 之间的值为：AUC：0.911，敏感性：90%，特异性：82.7%，PPV：74.2% 和 NPV：89.6%。我们开发了一个风险概率评分来预测 OPMD 患者是否存在 OC 或风险。我们建立了唾液 miRNA 特征，可以帮助诊断和预测 OC，彻底改变 OPMD 患者的治疗。总之，我们的结果为唾液 miRNA 管理 OC 以及使用源自唾液样本的 miRNA 的临床效用提供了新的思路。