Discarding negative values from input normalized bigwig tracks

[re2srm]

Hello,

I have chip-seq data for several histone marks for which I have created RPKM normalized bigwig tracks. I am now looking at the correlation between them using multiBigwigSummary and plotCorrelation functions.

My results show that some acetylation and methylation marks which are known to be mutually exclusive, show a low but positive correlation. Since they are expected to show a negative correlation, I think this positive correlation is a result of background signal. When I normalize the RPKM-bigwig tracks against RPKM-input track using bigwigcompare (using log2) and then plot correlation, I see better results with the marks showing a negative relationship.

However, if I look at the input normalized bigwig tracks in IGV or plot them with computeMatrix, a lot of regions show negative log2 ratios where the input has more signal then the chip. Since negative enrichment is not biologically meaningful, I am not sure if plots or correlations based on these tracks make sense or not.

My question is if it makes sense to discard negative values from the input normalized tracks before calculating correlation or plotting with computeMatrix and is there a way to do this in deeptools? (if not, do you have any advice on the best way to look at the correlation between different histone marks while incorporating the input sample in this case).

Thanks

[dpryan79]

That might be beneficial in this case. We don't have a way to do that built in, perhaps there's something in wiggletools that can filter a bigWig file?