Fig. 5. Gene set selection for healthy versus disease classification in multiple myeloma dataset.
Classification results of 40 genes selected by min-complexity strategy using 20 cells each iteration. a, Classification accuracy versus gene set size. b, The t-SNE plots of the entire filtered dataset. c, The t-SNE plots of the gene set selected with randomly sampling. d, The t-SNE plots of the gene set selected with `balanced' sampling. e, The expression level overlaid on t-SNE projection for genes selected.