Skip to main content
Fig. 3 | BMC Bioinformatics

Fig. 3

From: NEArender: an R package for functional interpretation of ‘omics’ data via network enrichment analysis

Fig. 3

Agreement between biological replicates in alternative approaches to differential expression analysis. a, c, e: Examples of Spearman rank correlations between DE analyses without replicates on gingivial epithelial versus tenocyte cells using samples from two different donor pairs (gingivial epithelial: #4 and #5; tenocyte: #2 and #3). Since p-values in the non-replicated DE analyses were not available, the plot A (RAW) represents raw fold change values for 16620 genes, while plots C and E represent p-values for 330 FGSs. In C and E, the vertical and horizontal brown dotted lines delineate p-values significant after the Bonferroni correction. Dashed grey line: the linear regression fit of Y on X (the R values shown in the corners were calculated using the rank formula and are thus independent of the fits). NEA and GSEA values were obtained for AGSs representing DE genes in each analysis that satisfied the criterion abs (log 2 (fold change))>2, i.e. the 4-fold change in either direction. These example rank R values from A, C, and E are plotted at respectively B, D, and F as big orange dots. Thus in B, D, and F, the Spearman coefficients from A, C, and E (top left corner) as well as from all other pairwise comparisons are plotted as a function of relative difference strength between respective transcriptomes (X axis). As an example, according to the results of fully powered DE analysis for plots A, C, and E, 13.8% genes were DE with adjusted p-value < 0.05. Hence the value 0.138 is used as X-coordinate for the orange dots. The grey dotted linear regression line and the Spearman rank R value quantify the relations between X and Y. Black and colored points correspond to 1) correlation values of non-replicated DE analyses with each other and 2) correlations where one of the two analyses was replicated, respectively. The boxplots G and H summarize results across all cell types and AGS versions for non-replicated vs. half-replicated options (1) and (2). All pairwise contrasts of GSEA and RAW against NEA were significant with p(H0) < 0.001. In Additional file 1: Figure S3, six plots for NEA and GSEA represent results of using all the six alternative AGS versions

Back to article page