Second-Order Polynomial Regularized Logistic Regression for Predicting Melanoma Patients Response to Immune Checkpoint Inhibitors

The incidence rate and mortality of melanoma continue to rise, becoming the focus of global health attention. Immunotherapy has shown great potential in the treatment of advanced or metastatic melano-ma. We proposed a second-order polynomial regularized logistic regression (SOPRLR) model for mining nonlinear interactions between features and applied it to a dataset of melanoma patients receiving anti PD-1 immune checkpoint inhibitors (ICI) treatment (n = 121). By comparing with five traditional machine learning models (RF, LR, SVM, LR-L, LR-R), the proposed model achieved highest F1 score in response to ICI. K-M survival analysis was conducted on the selected features, and the results showed significant differences in distinguishing responsive and non responsive patients.