Identification of stemness subtypes and features to improve endometrial cancer treatment using machine learning

Endometrial cancer is one of the most common malignant tumours in women, and cancer stem cells are known to play an important role in its growth, invasion, metastasis, and drug resistance. Immunotherapy for endometrial cancer is still under research. In this study, a total of 547 endometrial cancer cases were randomly divided into training set (351 cases) set and test set (196 cases). The stemness index of patients was calculated using the One-Class Logistic Regression (OCLR) machine learning algorithm to explore the clinicopathological differences between index levels. Stemness subtypes were determined according to the characteristics of cancer stemness and their clinicopathological characteristics, immune features, and therapeutic effects were described. Our study suggests that endometrial cancer is classified into two stemness subtypes. Stemness subtypes, which are associated with its clinical features, may be independent prognostic factors for endometrial cancer. The stemness subtypes differed significantly in immune activity, immune cell infiltration, and the immune microenvironment, including sensitivity to chemotherapeutic drugs and potential therapeutic compounds. Algorithms were utilised to construct a stemness subtype prediction model and predictor. These findings will provide guidance for the clinical diagnosis, treatment, and prognosis of endometrial cancer.

子宫内膜癌是女性最常见的恶性肿瘤之一，而癌症干细胞（cancer stem cells）已知在其生长、侵袭、转移及耐药过程中发挥重要作用。目前子宫内膜癌的免疫治疗仍处于研究阶段。本研究共纳入547例子宫内膜癌病例，按随机原则分为训练集（351例）与测试集（196例）。采用单类逻辑回归（One-Class Logistic Regression, OCLR）机器学习算法计算患者的干细胞干性指数，以探究不同指数水平间的临床病理差异。根据癌症干细胞干性特征确定干性亚型，并对各亚型的临床病理特征、免疫特性及治疗效应进行描述。本研究结果显示，子宫内膜癌可分为两种干性亚型。干性亚型与患者临床特征密切相关，或可作为子宫内膜癌的独立预后因素。不同干性亚型在免疫活性、免疫细胞浸润及免疫微环境方面均存在显著差异，包括对化疗药物和潜在治疗化合物的敏感性差异。本研究通过算法构建了干性亚型预测模型及预测工具。上述研究结果可为子宫内膜癌的临床诊断、治疗及预后评估提供指导。