基于血液自噬相关基因构建的静脉血栓栓塞机器学习模型研究

Study of machine learning model for venous thromboembolism based on blood levels of autophagy-related genes

目的利用生物信息学构建自噬相关基因(autophagy related genes,ATGs)的静脉血栓栓塞(venous thromboembolism,VTE)诊断预测模型。方法根据GSE19151数据集分析VTE人群的差异性ATGs(differentially expressed autophagy-related genes,DE-ATGs)和免疫细胞浸润,建立最优机器学习模型并通过诺谟图、校准曲线、决策曲线分析和外部数据集验证模型的预测效率。结果VTE患者中发现18个差异性表达的ATGs和差异的免疫细胞浸润。诺谟图、校准曲线和决策曲线分析证明机器学习模型对预测VTE具有一定的准确性,其中支持向量机器(support vector machine,SVM)机器学习模型对VTE的预测准确性最高。以SVM模型最重要的5个基因PRKCD、ULK1、CD46、PRKAR1A和FOS作为预测基因,在外部验证数据集表现出令人满意的性能,且与VTE患者的风险有关。结论研究首次揭示自噬与VTE之间的关系,并建立最优的机器学习模型来评估VTE患者的风险。

Objective To construct a diagnostic model based on autophagy-related genes(ATGs)for the clinical diagnosis and prediction of venous thromboembolism(VTE).Methods Based on GSE19151 dataset,differentially expressed ATGs(DE-ATGs)and infiltration of immune cell were analyzed in VTE patients.Optimal machine learning model was established and the predictive efficiency of the machine learning model was verified using a nomogram,calibration curve and decision curve analysis.Results Eighteen genes were identified as DE-ATGs.The nomogram,calibration curve and decision curve analysis also demonstrated high accuracy in predicting VTE.Support vector machine(SVM)machine learning model was found to best distinguish VTE patients.Using the top five most important genes from SVM model such as PRKCD,ULK1,CD46,PRKAR1A and FOS as predictor genes,SVM model exhibited satisfactory performance in an external validation dataset,and were significantly associated with the risk of VTE.Conclusion This study revealed the relationship between ATGs and VTE for the first time,and established an optimal machine learning model to evaluate the VTE patients and the risk of VTE.