Saturday, May 30, 2015
What's that DeltaCN thing do?
In PD 2.0 we see in every FDR box, even the once empty Fixed Value PSM validator, a line that says "Maximum Delta CN". Honestly, this has always been a little unclear to me, but we worked it out this week and then I thought of a good way of illustrating it.
While Sequest is going through matching theoretical fragmentations to your actual MS/MS spectra, it commonly comes up with more than one match. And it keeps them and gives them a score. The closest thing I could think of was the way that PepNovo+ does it. So I went and downloaded the newest version of the awesome and free de Novo GUI and did a quick run with their example data set.
(Click to expand). Now, this MS/MS spectra is exactly what I was looking for. It isn't an incredible spectra and the de Novo GUI came back with 4 possible sequence matches and the scores are pretty similar. The fist one is 43.98, the second is 41.48, and the last 2 are tied at 41.17. The delta CN is the degree of match between the scores of the possible peptide spectral matches (PSMs). For example, the default is 0.05. So we're saying if there are multiple matches we take the best one and anything that falls within 5% of that match. 5% less than 43.98 is 41.78. So, even though it looks like these three possibilities are pretty close we're only going to keep the very highest scoring one. If I changed it to 0.10, though, all 4 of these matches would be kept and it would be up to Percolator or target decoy to sort them out.
What is cool here is this: it is actually theoretically possible to get more peptides identified than you have MS/MS spectra. If there were lots of possibilities that were highly similar. If you manipulate the Delta CN cutoff you can actually ensure that. If you are using Scaffold to perform FDR on your Proteome Discoverer results, you actually want this to be the case. You set the Maximum Delta CN to 1 so every single match that Sequest makes goes into Scaffold and it does the rest.
Side note: I haven't downloaed the de Novo GUI in a while. And it keeps getting better. I know they were planning to add a BLAST function and they did! I can go to that match and just hit that little magnifying glass to the right and BOOM! it goes right to the BLAST interface and inputs my sequence into the BLAST bar. If you haven't checked this out you should.