How to normalize nucleotide coverage?

[asafl]

Hi,
I am interested in the function of a specific bacterial exonuclease. I want to know what are its gene targets and what are its processing sites. I have RNA-seq results of 2 biological replicates of WT bacteria and 3 biological replicates of the deleted RNAse.

I want to compare the bp coverage along genes to search for sites that have different coverage pattern between the WT and the mutant. But the mutant libraries have much more mapped reads than WT libraries. So what would be the best way to normalize the bp coverage between biological replicates and different samples (WT vs. mutant)?

Pay attention that I am not interested here in finding differentially expressed genes (which I already identified using DEseq), as a processed gene might not be considered as diff. expressed gene but still it may have different coverage pattern between WT and mutant.

Regards
Asaf

[jparsons]

If i understand correctly, you're not interested in expression level changes (number of transcripts) but rather only the changes in pattern of bp-level coverage, presumably due to the RNAse activity.

I'd consider just doing a rank-order correlation of counts across the first/last N nucleotides of every gene between treatments. Ideally, you'd see a be able to pick out anything with a significantly different pattern and reduced correlation, with no normalization necessary. Given how nasty (and variable) end-effects can be in RNA-seq, though, I'm not sure you'd be able to pick out small amounts of endonuclease activity.

I think the main thrust of my post is that you may not need to normalize the bp coverage, and if all you're doing is comparing the pattern of coverage, you likely don't need to. However, having looked at far too many bp-level coverage plots, I find it's convenient to normalize each gene to have an identical maximum value, for 'eyeballing'.

[jparsons]

I'd also be interested in any ideas for metrics for nt-level coverage "sameness", as that has been a long-term problem for me ; Every metric i've tried stumbles when it comes to coverage patterns that the eye quite clearly can tell are the 'same' but have, for example, one peak at 50% when that same peak is at 70% in the original.