基于正则化多项式回归的癌症患者免疫检查点阻断响应预测

Prediction of immune checkpoint blockade response of cancer patients based on regularized polynomial regression

探究特征之间的非线性相互作用关系,构建了用于预测癌症患者免疫检查点阻断响应的正则化多项式逻辑斯蒂回归模型.无进展生存期和总生存期的Kaplan-Meier曲线被用来筛选单个特征和联合特征,并据此构建了二次多项式判别函数.结合多项式负对数损失函数和弹性网络惩罚函数,提出了一种正则化多项式逻辑斯蒂回归模型,并通过特征拓维将其转化为线性模型求解.在泛癌、黑色素瘤、非小细胞肺癌和其他癌症数据集上与其他6种方法进行比较,结果表明所提方法取得了更高的免疫检查点阻断响应精度、F_(1)分数和AUC值.

This paper explored the nonlinear interaction between features and constructed a regularized polynomial logistic regression model for predicting immune checkpoint blockade response in cancer patients.The Kaplan-Meier curves for progression free survival and overall survival were used to screen for individual and combined features,a quadratic polynomial discriminant function was constructed.Combining the polynomial negative logarithmic loss function and the elastic network penalty function,a regularized polynomial logistic regression model is proposed,and it is transformed into a linear model through feature extension.Compared with other six methods for the data sets of pan cancer,Melanoma,non-small cell lung cancer and other cancers,the proposed method has achieved higher blocking response accuracy,F 1 score and AUC value of immune checkpoints.