Abstrakt: |
Background: Immune checkpoint inhibitors (ICIs) have revolutionized cancer treatment; however, a significant proportion of gastric cancer (GC) patients do not respond to this therapy. Consequently, there is an urgent need to elucidate the mechanisms underlying resistance to ICIs and identify robust biomarkers capable of predicting the response to ICIs at treatment initiation. Methods: In this study, we collected GC tissues from 28 patients prior to the administration of anti-programmed death 1 (PD-1) immunotherapy and conducted protein quantification using high-resolution mass spectrometry (MS). Subsequently, we analyzed differences in protein expression, pathways, and the tumor microenvironment (TME) between responders and nonresponders. Furthermore, we explored the potential of these differences as predictive indicators. Finally, using machine learning algorithms, we screened for biomarkers and constructed a predictive model. Results: Our proteomics-based analysis revealed that low activity in the complement and coagulation cascades pathway (CCCP) and a high abundance of activated CD8 T cells are positive signals corresponding to ICIs. By using machine learning, we successfully identified a set of 10 protein biomarkers, and the constructed model demonstrated excellent performance in predicting the response in an independent validation set (N = 14; area under the curve [AUC] = 0.959). Conclusion: In summary, our proteomic analyses unveiled unique potential biomarkers for predicting the response to PD-1 inhibitor immunotherapy in GC patients, which may provide the impetus for precision immunotherapy. [ABSTRACT FROM AUTHOR] |