Development and validation of post-Bz–based GEP80 predictive model. (A) Heat map of GEP80 expression levels 48 hours after Bz administration (top) and at baseline (bottom) in training set of 142 patients enrolled in the UARK2003-33 TT3A protocol. In both panels, samples are represented in columns and genes in rows. Each column represents one patient, and each row represents one gene. Samples are ordered by ascending GEP80(PB) score, which is indicated by the green bar on the top, and genes are ordered by hierarchical cluster analysis. The horizontal yellow line separates the unfavorable genes (above yellow line) from the favorable genes (below yellow line) in each panel. The vertical yellow line separates the low-risk (left of yellow line) from the high-risk patients (right of yellow line). (B) The GEP80(PB) model distinguishes between high-risk and low-risk patients in the training set, reflected in significantly different PFS (top) and OS (bottom). (C) Heat map of GEP80 expression levels 48 hours after Bz administration (top) and at baseline (bottom) in test set of 128 patients enrolled in the UARK2006-66 TT3B protocol. In both panels, samples are represented in columns and genes in rows. Each column represents one patient, and each row represents one gene. Samples are ordered by ascending GEP80(PB) score, which is indicated by the green bar on the top, and genes are ordered by hierarchical cluster analysis. The horizontal yellow line separates the unfavorable genes (above yellow line) from the favorable genes (below yellow line) in each panel. The vertical yellow line separates the low-risk (left of yellow line) from the high-risk patients (right of yellow line). (D) The GEP80(PB) model distinguishes between high-risk and low-risk patients in the test set, reflected in significantly different PFS (top) and OS (bottom).