Skip to main content
Figure 2 | BMC Bioinformatics

Figure 2

From: ChIPXpress: using publicly available gene expression data to improve ChIP-seq and ChIP-chip target gene ranking

Figure 2

Scatterplots of TF and target gene correlation and truncated correlation. Plots of the expression of mouse and human transcription factors (TF) against the expression of one of their known target genes (TG): (A) Oct4, and the expression of its target gene, Rrn3, (B) Gli3, and the expression of its target gene, Ezh2, (C) ESR1, and the expression of its target gene, COPZ1, and (D) TFAP2C, and the expression of its target gene, RAB3D, (E) Jarid2, and the expression of its target gene, Jund, and (F) Nanog, and the expression of its target gene, Fam168b. In (A,B,E,F), each point is one of 9,634 samples in the GPL1261 mouse expression compendium and in (C,D), each point is one of 18,257 samples in the GPL570 human expression compendium. (A-D) Oct4, Gli3, ESR1 and TFAP2C are examples of target gene activation by the TF, while (E-F) Jarid2 and Nanog are examples of target gene repression by the TF. The Pearson correlation (r) between each TF and TG increases for activated target genes and decreases for repressed target genes when selecting only the subset of samples with higher than average TF expression (blue). Furthermore, if one constructs a null correlation distribution for each TF by calculating the correlation between each gene in the compendium with the TF, one can show that standardized Z-values obtained by subtracting the mean and dividing by the standard deviation of the null correlation increase in magnitude when excluding samples in which the TF expression is below the average TF expression. This shows that there is increased separation between the observed correlation and the null distribution.

Back to article page