scGSECA:identification of altered biological processes in single-cell RNA-sequencing data by discretization of expression profiles

Bimodality of the distribution of expression levels has been observed in bulk RNA-seq. Two possible explanations underpin this kind of partitioning and refer to opposite paradigmatic models. At single-cell level, FISH experiments detected lowly and highly expressed genes in the same cells. We aimed at clarifying whether the bimodality observed in bulk RNA-seq data is a "herd effect" or a specific feature of expression profiles. Here we propose a cell-centric rather than gene-centric view to characterize expression profiles in single-cell RNA-seq (scRNA-seq) experiments produced by the high sensitivity protocol Smart-seq2. scGSECA is a high sensitivity tool for the identification of altered gene sets. Among the tested GSA methods scGSECA stands out especially in large cohorts of cells with subtle alterations with respect to competitor methods