Crux sequest-search for qexactive data

[Luminita]

Hello,

I have used Crux sequest-search with the parameters below to search a
sample run on a qexactive instrument.

precursor-window=10
precursor-window-type=ppm
isotopic-mass=mono
enzyme=trypsin
missed-cleavages=TRUE
num-decoys-per-target=0


Everything worked fine, but after post-processing with percolator I
obtained significantly less identifications compared to Mascot +
percolator.
Does anyone have any experience with using Crux for qexactive data?
And if yes, could I get some suggestions related to parameter values
to be used in crux?

Thank you in advance,
Luminita

[William Noble]

Hi Luminita,

I do not have experience with qexactive data, though I'd be happy to hear advice from anyone who does.

I would recommend starting by trying to figure out whether the problem lies in the search or in the calibration produced by Percolator.  To do so, rank your spectra according to Mascot+Percolator score and label each one according to whether the top-ranked candidate peptide is the same for Mascot and Crux.  You might want to do this first for a non-qexactive data set that you understand well, to establish a baseline.

• If the crux search is working well, then you should see that, for the high-scoring spectra, Mascot and Crux agree almost all the time.  If this is not the case, then you can examine some of the differences to try to understand why Crux is missing some good matches.  In particular, you should check whether the candidate peptides that Mascot identifies are even being considered by Crux.
• If, on the other hand, the two search engines produce largely the same top-scoring peptides for the high-ranked spectra, then the problem probably lies in the calibration produced by Percolator.  In that case, you should start looking into the feature values and see if there isn't something that needs to be tuned there.

Hope this helps.
Bill

[Sean McIlwain]

Luminita,

In addition, could you try adding to your parameter file for crux
sequest-search:

mz-bin-offset=0

and see if that fixes the problem?

Thanks,
Sean

> --
> You received this message because you are subscribed to the Google Groups
> "crux-users" group.
> To post to this group, send email to crux-...@googlegroups.com.
> To unsubscribe from this group, send email to
> crux-users+...@googlegroups.com.
> For more options, visit this group at
> http://groups.google.com/group/crux-users?hl=en.

--
Sean J. McIlwain, Ph. D.
Post-doctoral Fellow, Noble Lab,
Dept. of Genome Sciences,
University of Washington
http://noble.gs.washington.edu/~mcilwain

[Luminita Mihaela Moruz]

Hi Bill,

Thanks, I will definitely try to follow your advice and see if I can draw any conclusion there. My only observation when I looked at the data was that Crux+percolator was mostly performing worse at identifying shorter peptides (8-10 aa long). 

Thanks,
Luminita

[William Noble]

Luminita,

Mike MacCoss pointed out a potential confounder in this analysis.  Mascot will use the very high mass accuracy of the fragments measured in the qexactive, whereas Crux bins the data on the m/z axis. Crux would probably benefit from deisotoping the qexactive data using an approach like Hardklor first. Mascot's distiller program does this and also might explain the differences observed.

Bill

[Luminita Mihaela Moruz]

Hi Bill,

When I obtained the results mentioned, I did use Hardklör first. In any case, that was just one dataset. I will test some other ones, together with the parameter alteration suggested by Sean, and see if I obtain the same substantial difference between the two search engines.

Thanks,
Luminita