A pan-cancer analysis of the granzyme family based on the TCGA database and single-cell RNA sequencing data.

e14580 Background: Human Granzymes (GZMs) are mainly regarded as protective cytotoxic proteases that block the development of cancer. This study aimed to systematically analyze the expression signature, immune infiltration pattern, predictive values of prognosis and immune response, and the biological process associated with GZMs in different cancers. Methods: We estimated the predictive values of GZMs in prognosis and drug sensitivities by Cox regression analysis, immunophenoscore (IPS) algorithm, etc. We identified potential factors associated with the expression of GZMs, involving the somatic mutation and copy number variation of driver genes. We further investigate the distributions of GZMs and their discrepancy in single-cell resolution. The package CellChat and SCENIC were used for cell-cell communications and regulatory network reconstruction respectively. The ssGSEA algorithm was used to calculate the GZMs score for a gene set of five granzymes and the enrichment score (ES) of regulons based on the bulk-seq data. We chose the specific regulons to cluster patients in pan-cancer. Results: In the immunotherapy context, the higher GZMs score was associated with better overall survival (OS) in urothelial carcinoma (HR = 0.31, p< 0.01). In melanoma, the pooled HR value calculated by meta-analysis assay was 0.33 (95%CI: 0.189-0.58). In the single-cell resolution, we found the three patterns of GZMs expression in CD8+T/NK cells. Furthermore, we analyzed the discrepancy among different GZMs subtype cells in functions, cell-cell communications, and cell-type-specific regulons. In CD8+ Effector memory T cells, cluster 1 (GZMB+) and cluster 2 (GZMK+) exhibited different enriched pathways in adhesion, anti-bacteria, chemotaxis, migration, and metabolism. In the cell-cell communications analysis, the IFNG − (IFNGR1+IFNGR2) between CD8+ T cells and malignant cells was only significant in the C1 and C2 clusters. Pan-cancer patients were divided into five regulons clusters based on the regulons. Cluster 5 had the worst OS (HR = 8.49, 95%CI = 6.67-10.81, p< 0.01) compared to cluster 1 which had the best prognosis. Cluster 5 had higher ES of regulons including Myb, Hmgb2, E2f7, E2f8, and lower ES of Egr3. Conclusions: In conclusion, the expressions of GZMs, GZMs score, and the specific regulons identified in this study were associated with the prognosis in pan-cancer patients. GZMs score was a valuable biomarker that can predict the response to immunotherapy. The patients who belong to different regulons subtypes have specific transcription factors, which may be potential targets for the precision medicine of anti-tumor strategies in pan-cancer.