Manual validation of phosphopeptides

Search algorithms, post-searching processing, quantitation software, etc. Share and discuss software here.
JHunter
Glycine Member
Glycine Member
Posts: 5
Joined: Fri Feb 22, 2013 2:01 pm

Manual validation of phosphopeptides

Postby JHunter » Thu Mar 14, 2013 6:58 am

Hi,

Does anyone have any suggestions on how to manually validate phosphopeptides after database searching? I don’t feel comfortable completely trusting the results from our search databases…We are looking at sites on a protein complex after shotgun analysis, so I have ALOT of data to validate. Thanks!

J

User avatar
Doug
E. Coli Lysate Member
E. Coli Lysate Member
Posts: 307
Joined: Sun Jun 26, 2011 7:20 pm

Postby Doug » Thu Mar 14, 2013 7:40 am

Are you trying to validate the peptide ID itself or the location of the phosphorylation site?
- If your sample was enriched for phosphopeptides and most of your identified peptides were phosphorylated than your FDR should be reliable.

And are you using a site localization algorithm?
- Search algorithms often incorrectly assign the site of phosphorylation so this is highly recommended.

Frankly, I don't like the idea of manual validation at all. I think it is much more biased and subjective that using an algorithm. I would avoid it if possible.
"If we knew what we were doing it wouldn't be research." -AE

JHunter
Glycine Member
Glycine Member
Posts: 5
Joined: Fri Feb 22, 2013 2:01 pm

Postby JHunter » Thu Mar 14, 2013 7:54 am

We are trying to identify sites without any peptide enrichment, and validate the identification by using multiple digests with different enzymes to obtain a diverse sequence coverage, i.e trypsin, chymotrypsin, and GluC. I am using PhosphoRS in Proteome Discoverer as a site localization algorithm. We also use Scaffold to analyze our data manually and in some instances the algorithms (in Scaffold we can compare X!Tandem and Sequest data side-by-side) assign a high score, but the spectra is poor quality....advice?

justin
Phosphoserine Member
Phosphoserine Member
Posts: 11
Joined: Tue Jul 05, 2011 6:04 am

Postby justin » Thu Mar 14, 2013 7:38 pm

It is relatively difficult to manually validate phosphorylation for larger-scale analyses. For past publications, we've gone through and annotated each site that we report, but there is nothing fancy about it. We literally sat down and compared spectra from our phospho-IDs to the theoretical output from protein prospector (http://prospector.ucsf.edu/prospector/mshome.htm). Admittedly, this is tedious but it can be important if you don't feel comfortable with what you are seeing from PhosphoRS. Although most algorithms are good, none are perfect and incorrect assigments are somewhat common, as is tagging something as non-localized when there might be plenty of evidence for an assignment.

It is important to be wary of chimeric spectra (i.e., spectra that have ion fragments from more than one phospho-isoform). This happens a lot when you multiple potential phosphorylation sites on a single peptide because the different isoforms very frequently co-isolate (same mass and charge). If you're only interested in IDs, this is fine because you're just looking for evidence that a localized site exists. It becomes a serious problem if you're trying to quantify.

I couldn't agree more with Doug's comment about the bias introduced by manual validation. I think if you are going to do manual verification, it is important to very clearly layout the criteria that you use for making an assignment. Here is an example from a materials and methods section:

In all cases, product ions were required to be within 15 ppm of the theoretical m/z with a signal-to-noise ratio ≥ 3 to be used for sequence assignment.

Danny
Proton Member
Proton Member
Posts: 1
Joined: Tue Jul 09, 2013 9:00 am

Postby Danny » Tue Jul 09, 2013 9:27 am

Hi,

Further to the above discussion, I was wondering what people thought about my current set up for phosphoproteomic analysis of a subcellular organelle. I've been analysing samples using an Orbitrap elite (velos for MS/MS). I search the spectra using Mascot and use scaffold to view the phosphopeptides by using the phospho filter and applying scaffold's most stringent peptide score (95%), which typically gives a peptide FDR of <1%. I've then used spectral counts to compare phosphopeptide abundances across different samples (largely looking for things detected in one sample and not in another as a measure of relative abundance). As there are so many phosphopeptides to look at (more than 2000), I haven't done any manual validation yet. Does anyone have any experience/thoughts of whether Scaffold does a good job of correctly assigning phosphopeptides? I'm more interested in the peptide ID (i.e. my protein is phosphorylated in sample x more than sample y) than the location of the site, although both would be nice!
Thanks in advance


Return to “Bioinformatics”

Who is online

Users browsing this forum: No registered users and 1 guest