Finally a new blog post from me after my blog has been dormant for half a year - Lars Juhl Jensen

Thanks for the analysis Lars. I had a look at issue you pointed out in what it relates to our study http://pbeltrao.blogspot.com/2009... - Pedro Beltrao

It would be cool to know how much of the correlation you mention might be due to MS bias. - Pedro Beltrao

Thanks Pedro - redoing the same plot based on average number of phosphorylation sites per residue instead of on the raw counts was exactly what I wanted to do. It would just have been too much work to redo it without asking you for access to all your files. So I hoped that you would do it :) The result is as I expected; when I looked at protein lengths it did not explain the full signal, so there should be a correlation left after the length correction. - Lars Juhl Jensen

This is great discussion. One day, online papers will be wiki documents and you'll be able to edit one another's figures :-) - Neil Saunders

Intuitively, I would expect that due to biology the number of phosphorylation sites is directly proportional to the length of the protein (i.e. that the average density of phosphorylation sites is independent of protein length). When it comes to MS biases then I think that protein expression level is a much more important factor to take into account; a phosphorylation site on an abundant protein is much more likely to be picked up than one on a protein present in only one copy per cell. - Lars Juhl Jensen

Come to think of it: if there is a group of fairly small, highly expressed proteins (ribosomal proteins?), then this would explain the negative intercept with the y-axis in my plots. (Edit: rethinking the argument, this would give a positive intercept, not a negative one.) - Lars Juhl Jensen

Right; you can have small proteins with zero sites, but you can't have zero-length proteins. - Neil Saunders