The high expression group in each plot (blue) accounts for the upper quartile of expression levels for a particular transcript and the low expression group (red) the remaining 75%%.
To test this hypothesis, we split the putative ceRNAs into 4 groups by the quartiles of expression levels of miRPs (6.29, 6.88, and 7.52 in log2 scale; histogram in Additional file 3: Figure S3), resulting in 11,862,855 ceRNA pairs in each group.
To avoid bias in selection of cutoff point, we also re-analysed the results using median, instead of higher quartile, mRNA expression level as the cutoff point for Kaplan Meier analysis.
The mean expression of the 815 additional genes in post-GR samples was well below the lower quartile of the expression levels of genes common to both pre- and post-GR samples.
We plotted 5hmC density along the gene body for all genes separated in quartiles by their expression levels.
To visualize the prognostic impact, we performed Kaplan Meier analyses on the four quartiles based on expression levels.
Genes were then classified into quartiles based on their basal gene expression levels: 1st quartile is lowest and 4th is highest.
Interestingly, patients whose tumors were in the lowest quartile of Stat5a mRNA expression levels were associated with reduced time to breast cancer relapse, whereas the highest quartile of Stat5a mRNA expression levels associated with longer time to relapse.
When patients were divided into quartiles according to their expression levels of RRM2, 11 patients were in the lowest quartile (0.97 4.03), 20 patients in the intermediate quartiles (4.03 21.75) and 11 in the highest quartile (21.75 114.3) (Table 2).
In addition, HOXC9 was investigated separately by using quartile values of HOXC9 expression levels in a set of 244 neuroblastoma tumors.
