Refinement of Single-Cell RNA-seq Gene Expression Signatures with Combiroc

In this work we showcase how the combiroc R package can be used to refine and optimize gene signatures generated by high-throughput omics. Our workflow identifies subsets of gene expression signatures from traditional single-cell RNA sequencing experiments and allows distinguishing cells with higher confidence and fewer markers. Using a well-annotated single-cell RNA-seq dataset, we demonstrate that smaller signatures of lower-ranking genes selected with this combinatorial approach exhibit a greater power than top differentially expressed genes: these sub-signatures maintain a descriptive ability in terms of biological ontologies related to the cells to which they belong, defining actual and specific bio-semantic spaces.