SCMarker: Ab initio marker selection for single cell transcriptome profiling
Fig 1
Plotted as examples are (A) a bimodally distributed gene expression and (B) a unimodally distributed gene expression. From a binarized gene-cell expression matrix (C), a k-nearest co-occurrence neighbour (KNCON) graph and a k-nearest mutually exclusive neighbour (KNMEN) graph are constructed (D), based on which co- or mutually exclusively expressing gene pairs (CMEGPs) in the KNCON (node 1, 2 and 3, node 6 and 7, connected by red double arrows) and in the KNMEM (node 1, 2, 6 and 7, connected by green double arrows) can be identified. Marker genes (node 1, 2, 3, 6 and 7) are subsequently selected based on the CMEGPs.