Fig. 2

The performance of scSorter and four other methods on simulated and real datasets. a The performance of the five methods on simulated data of three scenarios. In the first scenario, we simulated ten cell types and the marker genes for each of these ten cell types were given. In this case, there were no “unknown” cell types in the expression data, and all known cell types appeared in the (scRNA-seq) expression data. This was the simplest scenario. In the second scenario, the marker genes for ten cell types were given but some of these ten cell types did not appear in the expression data. In the third scenario, there were ten cell types in the expression data but the marker genes of some of these ten cell types were unknown and thus cells from these cell types should be assigned to “unknown.” b The performance of the five methods on five real datasets