For example, I have samples sequenced both H3K27me3 and RNA-seq at 10 time point, and I want to see wether the expression level of genes is in negative correlation with H3K27me3 level. Is there a public accepted method? The way I adopt is that I treat the H3K27me3 sample as RNA-seq to calculate its CPM and the Pearson coefficient with CPM from RNA-seq. But this way completely ignore the peak concept which is important to ChIP-seq. Is there any better method?
•
written
by
ch8316f5eyu •
There nothing really wrong with what you're doing and I see ignoring the concept of peaks as a benefit. You might additionally do some clustering to see if there are subsets of genes that positively and others that negatively correlate (using deepTools, computeMatrix individually on the RNAseq and ChIPseq datasets, merge them together with computeMatrixOperations, and then cluster with plotHeatmap). But really, if it's obvious on the scatter plot of the CPMs then any method will show it.
Thanks for your reply. I am confused because I could not find any paper use this method. :)
Log in to answer this question.
