Deeptools correlation plot: between RNAseq and H3K27ac ChIPseq

Question

[aditi225]

Hi, I have been trying to plot correlation between my transcriptome and epigenome (H3K27ac) data but it shows almost no correlation. Within transcriptome and epigenome samples, the correlation is fine but I am not sure deeptools is a good way to plot correlation between RNAseq and H3K27ac ChIPseq. Here is my command:

python3 multiBamSummary bins --bamfiles /*R*bam  --smartLabels -o multibam.npz -p 20

python3 plotCorrelation -in multibam.npz --corMethod pearson --skipZeros --whatToPlot scatterplot -o multibam_perason.png   --outFileCorMatrix pearson.tab

Link to images:

[Devon Ryan]

Presumably you have MANY bins with no signal, have you tried excluding them?

[aditi225]

It remains the same even when I used BED files for locus specific correlations. Thanks

[Devon Ryan]

You might scroll around with IGV a bit just to eye-ball whether these values seem reasonable. I haven't looked at this particular correlation myself, so I don't really know what to expect. You may also use the --log1p option, which won't change the correlation, but will make the graphs easier to see.

[Gautier Richard]

I would recommend to call H3K27ac peaks, get nearest/overlapping genes with bedtools nearest, and plot the peaks maximum value alongside the associated genes expression with R / Python / Excel.

Otherwise, rank genes from the most expressed to the least expressed and make a heatmap with deepTools plotting H3K27ac at these genes TSS (computeMatrix followed by plotHeatmap) while keeping the expression ranked genes order (--sortRegions keep). I did that kind of analysis to validate RNA-seq / FAIRE-seq correlation (noisy old-fashioned open chromatin signal) and it worked like a charm.

[GenoMax]

Please use these directions: How to add images to a Biostars post

[aditi225]

I'll do that, thanks.