I wish the manuscript in the Journal of Proteome Research from Theo Luider and his team at the Erasmus Medical Center was available when I had to give my presentation at the London Biological Mass Spec Discussion Group. It’s got a great example of one of the advantages of our “quantify and then identify” analysis approach based on ion intensity measurements of peaks. I think I could have handed out this paper and then sat down!
An advantage of this approach, compared to one that relies on MS2 data and peptide identification, is how many peptide ions (including low abundant ones) you can view and use to quantify your sample.
One result in this paper showed Progenesis LC-MS revealed 23,654 MS signals in three replicate injections of one sample (incidentally this was a test of reproducibility where 82% of signals were found to have a CV <20%!). But only 6% of these MS signals could be identified as peptides from an NCBI database search and only 3.2% of these were peptide sequences relevant to the samples and found in all three injections.
If you’re only running one replicate injection in your LC-MS experiments you may want to look away now because only 1.8% of these relevant peptide sequences were found in one injection!
That was just looking at one sample. When you look at all seven donors Progenesis LC-MS showed 35,875 MS signals with 96% resemblance across all donors. The frequency of MS signals across all samples is more than ten times higher than the frequency of peptides assigned by an NCBInr search using MS2 data. The frequency of MS signals also covers a wider m/z range as you can see below. So if you rely on peptide identifications to quantify your label free LC-MS data who knows what you could be missing, especially if you run n = 1.
Frequency and distribution of MS signals (left) compared to MS2 based peptide identifications (right) from analysis of a complex sample. Note the y-axis scale is TEN TIMES higher on the left hand graph.
Progenesis LC-MS has been out in the world long enough for us to start seeing results published from it, so I’ll keep you posted with any highlights I find. In the meantime, try it on your own data and see how many more interesting features you can see?