Fig. 7

Application of PseudotimeDE, tradeSeq, and Monocle3-DE to the natural killer T cell dataset. a PCA visualization and inferred pseudotime by Slingshot. Pre-defined NKT subtypes are marked by colors. Slingshot returns a trifurcation topology, where the three lineages are NKT0 to NKT1, NKT0 to NKT17, and NKT0 to NKT2. b Histograms of all genes’ p-values in the three lineages calculated by the three DE methods. c Heatmaps of normalized enrichment scores (NESs, marked by colors) and their corresponding p-values (in numbers) from the GSEA. Each NES value and its corresponding p-value are calculated for each DE method and each lineage, based on the p-values of a DE method for a lineage and that lineage’s DE genes found from bulk RNA-seq data, denoted by “NKT1 bulk”, “NKT17 bulk,” or “NKT2 bulk” [37]. Note that among the three DE methods, PseudotimeDE outputs p-values that best agree with the lineage-specific DE genes from bulk data and thus most distinguish the three lineages. For instance, for the NKT1 lineage, PseudotimeDE’s small p-values are enriched in the “NKT1 bulk” gene set only, while tradeSeq and Monocle3-DE have small p-values enriched in at least two lineage-specific DE gene sets