Signature Definition Shapes Associations Between Exenatide-Related Transcriptional Programs and Macrophage Senescence Scores in Colorectal Cancer

Background: Treatment-derived expression signatures are often applied to untreated tumor cohorts, although their biological meaning may depend on signature construction and cellular composition. We examined the relationship between an exenatide-derived signature and senescence-associated macrophage expression in colorectal cancer. Methods: Archived pan-cancer and patient-level results were examined alongside a reconstruction from public mouse bulk RNA sequencing and colorectal single-cell matrices. Exenatide signatures were re-derived with tissue and batch adjustment, within-mouse correlation, and one-to-one orthologue mapping. Shared-gene removal, stricter macrophage annotation, expression-matched random signatures, virtual perturbation summaries, and three murine treatment comparisons assessed interpretation and robustness. Results: The original signature was associated with SenMayo in stomach, colon, and rectal tumor analyses. Reconstruction retained 209,207 colorectal cells and 4316 tumor macrophages, with 17 patients eligible for tumor-level analysis. Removing CXCR2 and IL1B from both signatures preserved a positive association (Spearman ρ = 0.804; adjusted β = 1.526; p = 0.00219). The corresponding mean expression association exceeded most expression-matched random signatures (empirical p = 0.006). Associations persisted with a blocked log-expression derivation but were absent with a blocked voom-derived signature using one-to-one orthologues (ρ = 0.056; p = 0.620). Archived virtual IL1B deletion and overexpression predicted opposing shifts after multiplicity adjustment. Treatment–macrophage fold-change correlations remained small (ρ = −0.038 to 0.046), with heterogeneous enrichment. The association was attenuated after adjustment for either inflammatory score—it remained statistically detectable after TNFα/NF-κB adjustment (β = 0.672, 95% CI 0.307–1.037) and was less precisely estimated after broader inflammatory response adjustment (β = 0.383, 95% CI −0.031–0.797). Conclusions: The original exenatide-derived signature was associated with SenMayo scores in colorectal tumor macrophages after removal of shared genes. This association overlapped with inflammatory expression and varied with signature derivation. Dependence on signature derivation limits its interpretation as a pharmacological response marker. These findings support investigation of inflammatory macrophage states without establishing drug-induced senescence.