Skip to main content
Fig. 8 | Genome Biology

Fig. 8

From: PseudotimeDE: inference of differential gene expression along cell pseudotime with well-calibrated p-values from single-cell RNA sequencing data

Fig. 8

Application of PseudotimeDE, tradeSeq, and Monocle3-DE to the cell cycle phase dataset. a Distributions of non-DE genes’ p-values by three DE methods with inferred pseudotime. Top: quantile-quantile plots that compare the empirical quantiles of non-DE genes’ p-values against the expected quantiles of the Uniform [0,1] distribution. Bottom: histograms of non-DE genes’ p-values. The p-values shown on top of histograms are from the Kolmogorov–Smirnov test under the null hypothesis that the distribution is Uniform [0,1]. The larger the p-value, the more uniform the distribution is. Among the three DE methods, PseudotimeDE’s p-values follow most closely the expected Uniform [0,1] distribution. b Distributions of DE genes’ p-values by three DE methods with inferred pseudotime. Top: scatter plots of DE genes’ p-values against the proportions of variance explained (PVE), which measure the strengths of genes’ inferred cyclic trends in the original study [38]. PseudotimeDE’s p-values (− log10 transformed) have the highest correlation with the PVE, indicating that PseudotimeDE identifies the genes with the strongest cyclic trends as the top DE genes. Bottom: histograms of all genes’ p-values. Blue and red colors represent the p-values of DE genes and non-DE genes (same as in (a) bottom), respectively. PseudotimeDE yields the best separation of the two gene groups’ p-values. c Quantile-quantile plots of the same p-values as in a on the negative log10 scale. PseudotimeDE returns the best-calibrated p-values. d FDPs of the three DE methods with the target FDR 0.05 (BH adjusted- p≤0.05). e ROC curves and AUROC values of the three DE methods. PseudotimeDE achieves the highest AUROC. f Power of the three DE methods under the FDP = 0.05 cutoff. PseudotimeDE achieves the highest power

Back to article page