Collagen-related Gene Expression Level Predicts the Prognosis and Immune Therapy Response

Background: Gastric cancer patients responded differently to the same treatment strategy and had various prognoses for the lack of biomarkers to guide the therapy choice.

Methods: RNA data of a local gastric cancer cohort with 103 patients were processed and used to explore potential treatment guiding factors. Cluster analysis was performed by non-negative matrix factorization. The expression level of collagen-related genes was evaluated by ssGSEA named collagen score (CS). Data from TCGA, ACRG, and an immune therapy cohort were utilized to explore prognosis and efficacy. Prognostic predictive power of CS was assessed using the nomogram.

Results: In our study, local RNA data were processed by cluster analysis, and it was found that cluster 2 contained a worse tumor infiltration status. The GSEA result showed that collagen-related pathways were differentially activated in two clusters. In TCGA and ACRG cohorts, the CS can be used as an independent prognostic factor (TCGA OS: p = 0.018, HR = 3.5; ACRG OS: p = 0.014, HR = 4.88). An immunotherapy cohort showed that the patients with higher CS had a significantly worse ORR (p = 0.0025). The high CS group contained several cell death pathways down-regulated and contained the worse tumor microenvironment. The nomogram demonstrated the survival prediction capability of collagen score.

Conclusion: CS was verified as an independent prognostic factor and potentially reflected the therapeutic effect of immunotherapy. The CS could provide a new way to evaluate the clinical prognosis and response information helping develop the collagen-targeted treatment.