基于机器学习的骨肉瘤双硫死亡基因诊断模型建立与分析

Construction and Analysis of Disulfidptosis Gene Diagnosis Model for Osteosarcoma Based on Machine Learning

目的通过机器学习算法鉴定与骨肉瘤(osteosarcima,OS)双硫死亡相关的特征基因,构建诊断预测模型,为进一步探索OS早期诊断的潜在生物学标志物和分子机制提供理论支持。方法差异表达分析用于鉴定OS双硫死亡差异表达基因(differential expression disulfidptosis-related gene,DE-DRG),通过最小绝对收缩选择算子(least absolute shrinkage and selection operator,LASSO)、支持向量机(support vector machines,SVM)和随机森林(random forest,RF)算法进一步鉴定具有诊断价值的OS双硫死亡特征基因,通过绘制受试者工作特征(receiver operating characteristic,ROC)曲线评估其诊断价值;同时构建评估疾病风险的列线图,并通过校准曲线和临床决策曲线评估列线图的有效性能。实时荧光定量聚合酶链反应(real-time quantitative polymerase chain reaction,RT-qPCR)检测特征基因在OS组织的表达量。结果共鉴定出2个具有较高诊断价值的OS双硫死亡特征基因(NDUFA11、RPN1),所构建的列线图对预测疾病风险具有较高的可靠性。RT-qPCR检测结果显示,在OS组织中,NDUFA11表达显著降低,而RPN1表达则显著升高(P<0.01)。结论本研究构建的OS双硫死亡基因诊断模型具有一定的诊断价值。

Objective To identify the characteristic genes associated with osteosarcoma(OS)disulfidptosis by machine learning algorithm,and to construct a diagnostic prediction model,so as to provide theoretical support for further exploring the potential biomarkers and molecular mechanisms of early diagnosis of OS.Methods Differential expression analysis was used to identify OS differential expression disulfidptosis-related genes(DE-DRG).The least absolute shrinkage and selection operator(LASSO),support vector machines(SVM)and random forest(RF)algorithms were used to further identify the diagnostically valuable OS disulfidptosis characteristic genes,and evaluated their diagnostic value by plotting the receiver operating characteristic(ROC)curve.At the same time,a nomogram was constructed to assess the risk of disease,and the effective performance of the nomogram was evaluated by calibration curve and clinical decision curve.The expression of characteristic genes in OS tissues was detected by real-time quantitative polymerase chain reaction(RT-qPCR).Results A total of two genes(NDUFA11,RPN1)were identified with high diagnostic value of characteristic genes associated with osteosarcoma(OS)disulfidptosis,and the nomogram constructed had high reliability for predicting disease risk.The results of RT-qPCR showed that NDUFA11 expression was significantly reduced,while RPN1 was significantly increased in OS tissue(P<0.01).Conclusion The established genetic diagnostic model of OS disulfidptosis in this study has certain diagnostic value.