Figure 5

Random forests survival analysis and Cox regression analysis to predict prognoses with the expression of inhibitory checkpoint genes in primary central nervous system lymphoma (PCNSL). (a) Variable importance derived from a random forests survival analysis. (b) Cox regression analysis for representative genes including programmed cell death 1 (PDCD1), CD274 (PD-L1), CD160, LGALS9, and CEACAM1. Akaike information criterion (AIC)-based optimization was performed, and hazard ratios with 95% confidence interval (CI) were shown. (c) Kaplan-Meier analysis on the expression levels of representative genes for inhibitory checkpoint. The patients were divided into the two subgroups with high (black line) and low (red line) expression by the cutoff score of the expression of the transcript variants, including programmed cell death 1 (PDCD1), programmed cell death ligand 2 (PDCD1LG2) (=PD-L2), CD80, LAG3, LGALS9, and CEACAM. Hazard ratio (HR) with 95% and p-value from log rank test were calculated. OS; overall survival.