Scholarlink

<p>This study aimed to develop a robust prediction model- using machine-learning algorithms based on the core indicators of the tumor immune microenvironment in gastric cancer patients to identify the predominant population likely to respond to immunotherapy. A total of 306 gastric cancer patients who received immunotherapy in our hospital from January 2022 to June 2024 were retrospectively included. They were randomly divided into a training set (<i>n</i> = 214) and a validation set (<i>n</i> = 92) at a ratio of 7:3. Five core predictors were screened from various clinical and pathological indicators: CD8⁺ T-cell density, Treg cell density, EBV infection status, TGF-β level, and CEA level. In the training set, indicators related to prognosis were screened by univariate analysis. After variable compression by Least Absolute Shrinkage and Selection Operator (LASSO) regression, independent influencing factors for poor prognosis were determined by multivariate logistic regression. Random Forest (RF), Gradient-Boosting Machine (GB), and K-Nearest Neighbor algorithm (KNN) prediction models were constructed using Python software. The model efficacy was evaluated by the area under the receiver operating characteristic curve (AUC), and the optimal model was selected. There was no statistically significant difference in the baseline data between the training set and the validation set patients (<i>P</i> &gt; 0.05). Multivariate logistic regression analysis showed that Treg cell density, TGF-β level, and CEA level were independent risk factors (<i>P</i> &lt; 0.05), while CD8⁺ T-cell density and EBV infection status (positive) were independent protective factors (<i>P</i> &lt; 0.05). The AUC values of the RF model (0.768 in the training set and 0.749 in the validation set) were significantly higher than those of the KNN (0.733, 0.662) and the GB model (0.718, 0.705), making it the optimal prediction model. This study successfully constructed and validated a prediction model integrating the core features of the tumor immune microenvironment. This model may assist in identifying gastric cancer patients who are likely to benefit from immunotherapy, and should be considered a complementary tool alongside existing biomarkers for clinical individualized decision-making.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

A prediction model based on tumor immune microenvironment for immunotherapy response in gastric cancer

  • 期刊论文

  • 出版模式：

    开放获取

  • 发表日期：

    2026年6月25日

查看全文
  • Xiangyu Xu,
  • Yingjie Song

摘要

This study aimed to develop a robust prediction model- using machine-learning algorithms based on the core indicators of the tumor immune microenvironment in gastric cancer patients to identify the predominant population likely to respond to immunotherapy. A total of 306 gastric cancer patients who received immunotherapy in our hospital from January 2022 to June 2024 were retrospectively included. They were randomly divided into a training set (n = 214) and a validation set (n = 92) at a ratio of 7:3. Five core predictors were screened from various clinical and pathological indicators: CD8⁺ T-cell density, Treg cell density, EBV infection status, TGF-β level, and CEA level. In the training set, indicators related to prognosis were screened by univariate analysis. After variable compression by Least Absolute Shrinkage and Selection Operator (LASSO) regression, independent influencing factors for poor prognosis were determined by multivariate logistic regression. Random Forest (RF), Gradient-Boosting Machine (GB), and K-Nearest Neighbor algorithm (KNN) prediction models were constructed using Python software. The model efficacy was evaluated by the area under the receiver operating characteristic curve (AUC), and the optimal model was selected. There was no statistically significant difference in the baseline data between the training set and the validation set patients (P > 0.05). Multivariate logistic regression analysis showed that Treg cell density, TGF-β level, and CEA level were independent risk factors (P < 0.05), while CD8⁺ T-cell density and EBV infection status (positive) were independent protective factors (P < 0.05). The AUC values of the RF model (0.768 in the training set and 0.749 in the validation set) were significantly higher than those of the KNN (0.733, 0.662) and the GB model (0.718, 0.705), making it the optimal prediction model. This study successfully constructed and validated a prediction model integrating the core features of the tumor immune microenvironment. This model may assist in identifying gastric cancer patients who are likely to benefit from immunotherapy, and should be considered a complementary tool alongside existing biomarkers for clinical individualized decision-making.