Fig. 4

Construction of the PFASRHSig survival risk model using multi-cohort transcriptomic data and machine learning. A Heatmap showing concordance index (C-index) of 101 machine learning models across six datasets: TCGA-LIHC (n = 355), CHCC (n = 159), LICA (n = 152), LIRI (n = 202), GSE14520 (n = 221), and GSE54236 (n = 78). B Top 10 predictive models by mean C-index. C Elastic net regression coefficient profiles across different values of log(λ). D Optimal λ selection using cross-validation with minimum partial likelihood deviance. (E-J) Kaplan–Meier survival curves comparing high- and low-risk groups (median cutoff) in six cohorts. Statistical significance was determined using the log-rank test