Construction of Prognostic Ferroptosis Signatures and Validation

Univariate Cox regression analysis was performed to examine the relationship between ferroptosis-related genes and OS in the TCGA UVM cohort. Moreover, two prognostic models were constructed, and the risk score was generated as follows: risk score = Expression_mRNA1 × Coefficient_mRNA1 + Expression_mRNA2 × Coefficient_mRNA2 +…Expression_mRNAn× Coefficient_mRNAn. The patients from TCGA and GEO databases were divided into high-risk and low-risk groups according to the median cutoff of the risk score. Next, to determine independent prognostic factors in UVM, univariate and multivariate Cox analyses were utilized to distinguish the clinicopathological parameters positively or negatively with the hazard ratio (HR). Subsequently, the Kaplan–Meier survival method was used to screen the availability of the prognostic model, and the receiver operating characteristic (ROC) curve was used to evaluate the prediction accuracy of 3-year and 5-year OS. To enhance prediction accuracy and interpretability, three GEO cohorts were further presented as validation sets to re-verify the prognostic model and selection of key ferroptosis modulators.