Metabolic programs associated with GDSC2-inferred oxaliplatin sensitivity are favorably prognostic in gastric cancer: transcriptomic and single-cell evidence.

Background: Oxaliplatin is widely used in gastric cancer (GC), yet molecular features associated with sensitivity are heterogeneous and incompletely defined. Pharmacogenomic models trained in cancer cell lines can generate hypotheses about drug-response biology, but their transfer to patient tumors requires cautious interpretation. We, therefore, integrated bulk transcriptomic, pharmacogenomic, prognostic, and single-cell data to characterize transcriptional programs associated with GDSC2-inferred oxaliplatin sensitivity in GC. Methods: A ridge-regression model trained on 396 GDSC2 cell lines was used to predict oxaliplatin area-above-curve (AAC) from expression data and was transferred to 412 TCGA-STAD primary tumors. Resistant-like and sensitive-like tumors were defined by the lower and upper predicted-AAC quartiles. Differential expression, gene-set enrichment, WGCNA, immune-cell scoring, metabolic-program scoring, and ROC analyses were performed, respectively. Prognostic associations were examined in GSE62254 and GSE26253, and cell-type localization was explored in the GSE134520 single-cell dataset. For experimental corroboration, AGS gastric cancer cells were exposed to oxaliplatin (10-20 μM) for 24 h, followed by qRT-PCR analysis of ALB, LDHA, FASN, and CPT1A mRNA expression relative to untreated AGS controls. Results: Cross-validation of the GDSC2 ridge model showed moderate agreement between predicted and observed oxaliplatin AAC (Pearson r = 0.50; Spearman rho = 0.44). In TCGA-STAD, resistant-like tumors showed higher glycolysis scores (Mann-Whitney Conclusion: Gastric tumors classified as oxaliplatin resistant-like by a GDSC2-derived transfer model exhibit coordinated metabolic and immune transcriptional features, particularly increased glycolytic and oxidative programs. However, these findings describe model-inferred drug sensitivity rather than clinically observed oxaliplatin resistance, and the favorable prognostic associations of several resistance-associated metabolic programs emphasize the distinction between drug-sensitivity inference and prognosis. The AGS cell qRT-PCR experiment provides limited mRNA-level experimental corroboration for selected metabolic genes under oxaliplatin exposure but does not establish protein-level or causal mechanisms. Direct validation in clinically annotated oxaliplatin-treated cohorts and additional functional experiments are required.