CKD基因预测模型优化策略

Currently, using both WGCNA and random forest algorithms to screen for core genes associated with CKD progression, we have only achieved an optimal prediction accuracy of 0.687. The validation set GSE60861 includes 20 patients with progressive disease and 52 stable patients, with a balanced patient composition. To address the lower predictive performance of this gene set, we can incorporate feature genes that contribute more significantly to CKD progression, identified by other algorithms such as SHAP, into the prediction model. This may further enhance the characteristic features of the gene set.