Fig. 5

Construction and validation of ECMGPS based on machine learning. A C-index of 101 prediction models using 10 machine learning algorithms across five cohorts. B Coefficients of 21 model genes obtained from 51 prognostic ECMGs using the CoxBoost algorithm. C Correlation network among the 21 model genes in PCa. Line thickness represents the strength of association, and color indicates the direction of the association. Dot size reflects the effect of each gene on prognosis, while color denotes gene expression. D PCA and t-SNE plots showing the distribution of low- and high-risk groups. E Kaplan–Meier curves showing bRFS in the low- and high-risk groups. F Receiver operating characteristic (ROC) curves showing 1-, 3-, and 5-year bRFS in the low- and high-risk groups