Molecular classification reveals the diverse genetic and prognostic features of gastric cancer: A multi-omics consensus ensemble clustering.

BACKGROUND

Globally, gastric cancer (GC) is the fifth most common tumor. It is necessary to identify novel molecular subtypes to guide patient selection for specific target therapeutic benefits.

METHODS

Multi-omics data, including transcriptomics RNA-sequencing (mRNA, LncRNA, miRNA), DNA methylation, and gene mutations in the TCGA-STAD cohort were used for the clustering. Ten classical clustering algorithms were executed to recognize patients with different molecular features using the "MOVICS" package in R. The activated signaling pathways were evaluated using the single-sample gene set enrichment analysis. The differential distribution of gene mutations, copy number alterations, and tumor mutation burden was compared, and potential responses to immunotherapy and chemotherapy were also assessed.

RESULTS

Two molecular subtypes (CS1 and CS2) were recognized by ten clustering algorithms with consensus ensembles. Patients in the CS1 group had a shorter average overall survival time (28.5 vs. 68.9 months, P = 0.016), and progression-free survival (19.0 vs. 63.9 months, P = 0.008) as compared to those in the CS2 group. Extracellular associated biological process activation was higher in the CS1 group, while the CS2 group displayed the enhanced activation of cell cycle-associated pathways. Significantly higher total mutation numbers and neoantigens were observed in the CS2 group, along with specific mutations in TTN, MUC16, and ARID1A. Higher infiltration of immunocytes was also observed in the CS2 group, reflective of the potential immunotherapeutic benefits. Moreover, the CS2 group could also respond to 5-fluorouracil, cisplatin, and paclitaxel. The similar diversity in clinical outcomes between CS1 and CS2 groups was successfully validated in the external cohorts, GSE62254, GSE26253, GSE15459, and GSE84437.

CONCLUSION

The findings provided novel insights into the GC subtypes through integrative analysis of five -omics data by ten clustering algorithms. These could provide potential clinical therapeutic targets based on the specific molecular features.