how to do average profiling only with bases covered in wiggle?

Question

[chxu02]

My wiggle/bedgraph records the DNA methylation level of CpGs (from 0-100). While I do average profiling around TSS with CEAS, the curve is all the way close to the x-axis. I bet CEAS is taking every unmapped base with a "0" value, bringing down the whole curve to 0. Is there any other way to do this job?

Answer 1

[Alex Reynolds]

Use wig2bed to convert wiggle to BED, and then bedops --not-element-of 1 to exclude methylation signal regions that are unmapped (or, conversely, bedops --element-of 1 for signal over mapped regions). Use the filtered methylation signal file with bedmap --mean to calculate mean signal over regions of interest.

[chxu02]

Hi Alex, I read the document you provided and found that bedmap --mean actually calculates the average signal along each interval in the reference file. What I'm going to do is to calculate the average signal across the same bin from multiple intervals specified in the reference file (eg., TSS+-3kb), to get an average profile of the signal along multiple TSS. Any further suggestion?

[Alex Reynolds]

If you first need to generate expanded intervals before taking the mean, you can use bedops --everything --range N to generate a set of symmetrically-padded reference intervals. Then you could bedmap --mean and aggregate results in R.