Fig. 6
From: A rank-based marker selection method for high throughput scRNA-seq data
Error rates of both the nearest centroids classifier (NCC; (a) and (b)) and the random forests classifier (RFC; (c) and (d)) on the Paul data set. Figure (b) (respectively (d)) is a detailed image of the error rate of the different methods using the NCC (respectively RFC) when smaller numbers of markers are selected. Figure (b) details up to 220 total markers to make clear how similar the methods perform when small numbers of markers are selected. Figure (d) examines up to 350 total markers to detail the performance of the methods when small numbers of markers are selected as well as get an idea for the increasing behavior and noisy nature of the curves