Efficient Identification of Co-fragmented Peptides

Publish a paper recently? Share it here. Or discuss other proteomics-related publications.
Zhang
Angiotensin Member
Angiotensin Member
Posts: 27
Joined: Fri Aug 24, 2012 8:08 am

Efficient Identification of Co-fragmented Peptides

Postby Zhang » Sat Aug 09, 2014 2:56 pm

Hi everybody,

We recently published our findings of identifying co-fragmented precursors, by simply searching the identical MS/MS spectra with alternative precursor information. The workflow is inspired by the very popular MaxQuant workflow, but developed based on an integration of several free proteomics tool-sets: :) OpenMS/TOPP, ProteoWizard, Morpheus, COMPASS, Pyteomics, etc.; and plus some pre- and post-processing in-house Python scripts.

The performance is surprisingly good when analyzing high-resolution Orbitrap data with a ppm-level mass accuracy. Although the re-formulated Morpheus scoring method used in the workflow had been heavily criticized by reviews, I think the topic of super simple search engine is worth discussing in this high-resolution era.

As a result, with 'cloning' MS/MS spectra for co-selected precursors, at 1% FDR we matched on average more than one peptides out of each MS/MS event, broke the conventional paradigm of 'one MS/MS ~ one peptide ID' for data-dependent acquisition (DDA). This workflow utilizes super accuracy from Orbitrap, and enables efficiently analyzing around 5000 human proteins with single-shot LC-MS/MS.



Paper:
http://www.ncbi.nlm.nih.gov/pubmed/25100859
Zhang, B., Pirmoradian, M., Chernobrovkin, A., & Zubarev, R. a. (2014). DeMix Workflow for Efficient Identification of Co-fragmented Peptides in High Resolution Data-dependent Tandem Mass Spectrometry. Molecular & Cellular Proteomics : MCP. doi:10.1074/mcp.O114.038877


Link to the codes (sorry for my programming skills 囧 ;) :
https://github.com/userbz/DeMix

I am willing to have your questions, comments and criticisms.


P.S. Special thanks to Craig.

Omics
Phosphoserine Member
Phosphoserine Member
Posts: 14
Joined: Sat Oct 06, 2012 10:34 am

Postby Omics » Wed Aug 13, 2014 1:28 am

For Mascot, the cutoff of E-value<0.05 is used for comparison. However, based on my experiences, peptides with a Mascot score>15 generally have a FDR of <1% for high-resolution Q Exactive data. When E-value<0.05, the Mascot score is generally above 30 or 40 depending on the size of database and parameters used. Thus, it is not a fair comparison for Mascot.

As shown in Table 2, the increase of unique peptide identification is negligible (23133 to 23095) for Mascot with the use of DeMix (possibly due to the high cutoff), indicating that the increase for other database search engine is possibly mostly low-score ones. Have you checked the difference of score between the first and second matched peptides?

Zhang
Angiotensin Member
Angiotensin Member
Posts: 27
Joined: Fri Aug 24, 2012 8:08 am

Postby Zhang » Wed Aug 13, 2014 3:01 am

Hi Omics,

Yes, E<0.05 is a very strict filter, which seemed not quite fair in this comparison. However, it was a conventional standard before the target-decoy strategy has been widely accepted. And, indeed, if use 1%FDR to trim Mascot result, you will see a similar improvement coming from precursor-deconvolution. It shows that chimeric spectra are generally scored lower by Mascot due to the non-chimeric presumption in its scoring model. It is also the reason we used the type of ion-counting method in peptide identification.

We have received a lot of criticisms on using the scoring method. Some people (for example: http://thegpm.org/basics/ ) just hate this method, even if it is simple, straightforward, accurate and fast. But anyway, it is a free choice, actually any search engine can be used in this workflow. Just keep in mind that sometimes PSMs are scored too low because they have too many extra peaks or assigned with wrong precursors; and over 90% of the MS/MS spectra have extra peaks from co-selected precursors in single-shot DDA experiments.

--
Bob


Return to “Literature”

Who is online

Users browsing this forum: No registered users and 1 guest