Abstract
We proposed gdGSE, a novel computational framework for gene set enrichment analysis. Unlike conventional methods that rely on continuous gene expression values, gdGSE employs discretized gene expression profiles to assess pathway activity. This approach effectively mitigates discrepancies caused by data distributions. This algorithm consists of two steps: (1) applying statistical thresholds binarizing gene expression matrix, and (2) converting the binarized gene expression matrix into a gene set enrichment matrix. Our results demonstrated that gdGSE could robustly extract biological insights from a diverse array of simulated and real bulk or single-cell gene expression datasets. Notably, gene set enrichment scores by gdGSE exhibited enhanced utility in downstream applications: (1) precise quantification of cancer stemness with significant prognostic relevance; (2) enhanced clustering performance in stratifying tumor subtypes with distinct prognoses; and (3) more accurate identification of cell types. Remarkably, the pathway activity scores by gdGSE showed > 90 % concordance with experimentally validated drug mechanisms in patients-derived xenografts and estrogen receptor-positive breast cancer cell lines. Our algorithm proposes that discretizing gene expression values provides an alternative method for evaluating pathway enrichment, applicable to both bulk and single-cell data analysis.