Abstract
Identifying gene expression programs underlying cell-type identity as well as cellular activities (e.g. life-cycle processes, responses to environmental cues) is crucial for understanding the organization of cells and tissues. Although single-cell RNA-Seq (scRNA-Seq) can quantify transcripts in individual cells, each cell’s expression profile may be a mixture of both types of programs, making them difficult to disentangle. Here, we develop an adapted non-negative matrix factorization approach, consensus NMF (cNMF), as a solution to this problem. We rigorously benchmark it against existing and novel scRNA-Seq methods, and cNMF performs best, increasing the accuracy of cell-type identification while simultaneously inferring interpretable cellular activity programs in scRNA-Seq data. Applied to published brain organoid and visual cortex scRNA-Seq datasets, cNMF refines the hierarchy of cell-types and identifies both expected (e.g. cell-cycle and hypoxia) and intriguing novel activity programs. We make cNMF and related tools available to the community and illustrate how this approach can provide key insights into gene expression variation within and between cell-types.