1. The reported number of significant H3K27ac regions in Figure 2 appears inconsistent with the Results. Figure 2A–B identifies 848 regions meeting the stated FDR and fold-change criteria, whereas the text reports 682 regions, comprising 657 non-responder-enriched and 25 responder-enriched regions. Could the authors clarify which result is correct and provide the corresponding DiffBind output and analysis code?
2. The PDX validation in Figure 3D appears to analyze 246 mouse-level observations by one-way ANOVA even though the strong/weak H3K27ac classification was assigned to only 15 PDX models. Because multiple mice derived from the same PDX model may not represent independent tests of the signature, could the authors provide a model-level or hierarchical analysis accounting for PDX identity? It would also be helpful to show whether the clinical signature is reproduced using only pretreatment biopsies, since the discovery comparison appears to include six post-treatment biopsies collected upon progression.