Single-cell mean rank gene set scoring method for between-dataset comparison of scRNA-seq data

Abstract

The surge in single-cell RNA sequencing (scRNA-seq) data offers a unique chance for researchers to understand functional changes in biological processes and diseases through gene set scoring across diverse datasets. Despite this, current methods for comparing scRNA-seq data at the signaling pathway level across datasets remain untested. To bridge this gap, we introduce the single-cell mean rank gene set scoring (scMRGSS) method, which assesses gene set activity between different scRNA-seq datasets. Leveraging gene expression ranks within each dataset, scMRGSS calculates mean rank scores for gene sets, enabling the comparison of their relative enrichment or depletion across datasets. Demonstrating its efficacy through simulated and real datasets, scMRGSS proves to be a simple yet informative tool for comparing gene set activity between cell types across diverse datasets. Its robustness against sequencing depth and dropout rate variations underscores its value for integrative scRNA-seq data analysis. Applying the method, we uncover that abnormal activity in oxidative phosphorylation and NF-κB signaling pathways in glioblastoma cancer cells may not solely stem from neurodevelopmental programs. Notably, the highest activity of these pathways is observed in the mesenchymal cancer cell type, emphasizing the need to target specific cell types in glioblastoma drug development.