Utilizing machine learning to integrate single-cell and bulk RNA sequencing data for constructing and validating a novel cell adhesion molecules related prognostic model in gastric cancer

Abstract

Background: Cell adhesion molecules (CAMs) play a vital role in cell-cell interactions, immune response modulation, and tumor cell migration. However, the unique role of CAMs in gastric cancer (GC) remains largely unexplored.

Methods: This study characterized the genetic alterations and mRNA expression of CAMs. The role of CD34, a representative molecule, was validated in 375 GC tissues. The activity of the CAM pathway was further tested using single-cell and bulk characterization. Next, data from 839 patients with GC from three cohorts was analyzed using univariate Cox and random survival forest methods to develop and validate a CAM-related prognostic model.

Results: Most CAM-related genes exhibited multi-omics alterations and were associated with clinical outcomes. There was a strong correlation between increased CD34 expression and advanced clinical staging (P = 0.026), extensive vascular infiltration (P = 0.003), and unfavorable prognosis (Log-rank P = 0.022). CD34 expression was also found to be associated with postoperative chemotherapy and tumor immunotherapy response. Furthermore, the CAM pathway was significantly activated and mediated poor prognosis. Additionally, eight prognostic signature genes (PSGs) were identified in the training cohort. There was a substantial upregulation of the expression of immune checkpoints and a pronounced infiltration of immune cells in GC tissues with high PSG score, which is consistent with the prediction of increased sensitivity to immunotherapy. Moreover, 9 compounds from the CTRPv2 database and 13 from the Profiling Relative Inhibition Simultaneously in Mixture (PRISM) database were identified as potential therapeutic drugs for patients with GC with high PSG score.

Conclusion: Thorough understanding of CAM pathways regulation and the innovative PSG score model hold significant implications for medical diagnosis, potentially enhancing personalized treatment strategies and improving patient outcomes in GC management.

Keywords: Cell adhesion molecules; Gastric cancer; Machine learning; Medical diagnosis; Therapeutic target.