卵巢透明细胞癌（OCCC）是上皮性卵巢癌的一种独特亚型。晚期 OCCC 预后较差。因此，我们的目标是对精准医疗进行风险分层。我们对 44 名处于 Figo II-IV 期的 OCCC 患者进行了大型下一代测序 (NGS) 基因检测。然后，通过机器学习算法，包括极端梯度提升（XGBoost）、随机生存森林（RSF）和Cox回归，我们筛选了与预后相关的特征基因，并构建了用于风险分层的5基因panel。通过受试者工作特征曲线和决策曲线分析，将5基因panel的预测效果与FIGO分期和残留疾病进行比较。通过机器学习算法筛选出与预后相关的特征突变基因，包括MUC16、ATM、NOTCH3、KMT2A、和CTNNA1。 5基因panel可以有效区分内部和外部队列中晚期OCCC的预后以及铂类反应，预测能力优于FIGO分期和残留病。基因突变，包括MUC16、ATM、NOTCH3、 KMT2A 和 CTNNA1 与晚期 OCCC 的不良预后相关。根据这些基因进行的风险分层显示出可接受的预后和铂类反应预测能力，表明有可能成为精准医学的新靶标。版权所有 © 2024 Elsevier Inc. 保留所有权利。

Ovarian clear cell carcinoma (OCCC) is a unique subtype of epithelial ovarian cancer. Advanced OCCC display a poor prognosis. Therefore, we aimed to make risk stratification for precise medicine.We performed a large next generation sequencing (NGS) gene panel on 44 patients with OCCC in FIGO stage II-IV. Then, by machine learning algorithms, including extreme gradient boosting (XGBoost), random survival forest (RSF), and Cox regression, we screened for feature genes associated with prognosis and constructed a 5-gene panel for risk stratification. The prediction efficacy of the 5-gene panel was compared with FIGO stage and residual disease by receiver operating characteristic curve and decision curve analysis.The feature mutated genes related to prognosis, selected by machine learning algorithms, include MUC16, ATM, NOTCH3, KMT2A, and CTNNA1. The 5-gene panel can effectively distinguish the prognosis, as well as platinum response, of advanced OCCC in both internal and external cohorts, with the predictive capability superior to FIGO stage and residual disease.Mutations in genes, including MUC16, ATM, NOTCH3, KMT2A, and CTNNA1, were associated with the poor prognosis of advanced OCCC. The risk stratification according to these genes demonstrated acceptable prediction power of prognosis and platinum response, suggesting the potential to be a novel target for precision medicine.Copyright © 2024 Elsevier Inc. All rights reserved.