A T-GEP–based MetSig related to oxidative metabolism is significantly associated with outcome in DLBCL. (A) Heatmap showing the expression levels of the 6 genes representing the MetSig selected based on supervised PFS analysis. Each row corresponds to 1 Z score normalized gene expression levels and each column corresponds to 1 patient. The expression change from left to right corresponds to the MetSig stratification. (B) Gene ontology analysis showing the cellular component involved in the T-GEP–based metabolic stratification. The node color changes from red to blue in descending order according to the adjusted P values. The size of the node represents the number of counts. (C) Kaplan-Meier curve for PFS in patients with DLBCL of the discovery cohort. Patients were divided into 2 groups, MetSig-high vs MetSig-low, by applying the maximally selected rank statistics. MetSig-high patients had significant worse outcome compared with MetSig-low patients. (D) OS of the 48 patients in the discovery cohort according to the MetSig showing significant differences in outcome between MetSig-low and MetSig-high patient subsets. (E) Forest plot depicting multivariable analysis for PFS (discovery cohort). According to this analysis, only the COO as determined by NanoString-based T-GEP (COO Nano) and the MetSig retained statistical significance for PFS. AIC, Akaike information criterion; NADPH, nicotinamide adenine phosphate; ref, reference; TCA, tricarboxylic acid.