Fig. 4
From: Robust principal component analysis for accurate outlier sample detection in RNA-Seq data
Comparison of DEG detection performance of the eight different data analysis strategies. a Number of DEGs; b Overlap of DEGs for the four strategies before outlier removal; c Overlap of DEGs for the four strategies after outlier removal; d Volcano plot of gene expression changes using S5. The x axis specifies the log2 fold-changes, and the y axis specifies the negative logarithm to the base 10 of the adjusted p values. Blue dots represent genes expressed at significantly higher (n = 792) or lower (n = 600) levels upon SnoN loss, respectively (adjusted p < 0.1). DEGs identified by S1(2), DEGs identified by S2 but not S1 (1), DEGs identified by S3 but not S1, S2 (9) and DEGs unique to S4 (8) and DEGs identified by S5 were labeled with respective colors; e Volcano plot of gene expression changes using S5 as described in panel (d); DEGs identified by S5, DEGs unique to S6 (29), DEGs identified by S7 but not S8 (6) and DEGs unique to S8 (8) were labeled with respective colors. Using qRT-PCR results as the reference standard we compared (f) sensitivity, specificity, positive predictive value (PPV), negative predictive value (NPV), PCR concordance, overall agreement between DESeq2 and qRT-PCR results and (g) false positive rate (FPR) and false negative rate (FNR) of the eight strategies. h Selective gene ontology (GO) terms from GO enrichment analyses of genes that determined to be up-regulated by S5 (S5Up), S6 (S6Up) or down-regulated by S5 (S5Down) and S6 (S6Down) in conditional SnoN KO samples. Before outlier removal: S1: Strategy 1; S2: Strategy 2; S3: Strategy 3; S4: Strategy 4; After outlier removal: S5: Strategy 5; S6: Strategy 6; S7: Strategy 7; S8: Strategy 8