Fig. 3

Lasso regression was used to screen the MRS for the predictive model; a a bar plot shows 32 mutant genes significantly associated with OS; b Lasso coefficient profiles of mutated genes in our iCCA cohort. c A coefficient profile plot was generated to find the optimal parameter (lambda). d Hazard ratios (HR) for the selected prognostic mutated genes were generated using multivariate Cox analysis and are shown in Forest plots; e KM plotter shows differences between the high- and low-MRS groups, indicating that high MRS was associated with poor survival outcomes. f Time-dependent ROC curve analysis shows that the AUCs for MRS were 0.759, 0.732, and 0.728 for 1, 3, and 5-year OS, respectively; this analysis demonstrates the satisfactory predictive accuracy of the MRS model