Fig. 1

Overview and validation of CellMapper. a Schematic of the approach. CellMapper takes as input a cell type-specific query gene (green) and a set of gene expression data and finds genes with a similar expression profile to the query gene (e.g. “Gene C” above, yellow profile). b Performance comparison between CellMapper and the machine learning algorithm, in silico nano-dissection [17]. CellMapper and in silico nano-dissection were each applied to identify podocyte genes and evaluated based on the recovery of an experimentally defined set of podocyte genes [23]. In silico nano-dissection was applied using the training set selected by Ju et al. [17] for their analysis (46 query genes and 97 negative control genes) or a smaller training set of ten query genes and ten negative control genes (the smallest training set permitted by the algorithm, see “Methods”). CellMapper identified the experimentally defined podocyte genes with similar precision to in silico nano-dissection at all levels of recall, despite using only one query gene