Skip to main content
Fig. 5 | BMC Bioinformatics

Fig. 5

From: eSVD-DE: cohort-wide differential expression in single-cell RNA-seq data using exponential-family embeddings

Fig. 5

A Number of case and control individuals and the total number of cells across cell types for the Velmeshev dataset, analogous to Fig. 4A. B Volcano plot showing eSVD-DE’s results for layer 2/3 cells, where the set of genes exceeding the empirical FDR cutoff are depicted in orange. The SFARI, bulk DE, and housekeeping genes are in purple, blue, and green, respectively. The Fisher exact p-value between the selected genes via eSVD-DE and the bulk DE genes is noted. C Comparison of the eSVD-DE p-values against the DESeq2 p-values, and the genes exceeding an FDR cutoff for DESeq2 are shown in yellow. The correlation between the two sets of negative \(\log _{10}\) p-values is noted. D Increase in enrichment for relevant GO terms among the 331 DE genes found by eSVD-DE, the 144 DE genes found by DESeq2, and the 109 genes initially reported by the authors. E Upset plot comparing the SFARI or bulk DE genes with the top 100 DE genes estimated by eSVD-DE, DESeq2, MAST, or SCTransform. F Downsampling experiment, illustrating the stability of the 50 genes with the largest test statistics in magnitude among the SFARI or the bulk DE genes as the dataset is artificially downsampled. The values on the y-axis is the ratio of the test statistic between these genes and the housekeeping genes

Back to article page