Dear all,
when we run PeptideProphet on pepXML Mascot DAT files, we always
notice an 'odd' peak on the f-values charts (see link 1).
It is a sharp peak positioned at negative f-values, somewhere in
between the positions where one would expect the Gamma and Normal
distributions of the negative and positive models.
It is always present in the charts for charge +2 and often visible in
the charts for charge +3.
I have never seen such a peak in the distributions presented in
Peptide Prophet papers or presentations.
And I dont' see it if I repeat the search with xTandem (see link 2).

I have checked the f-values in Excel and I have noticed that almost 5%
of the PSM are assigned exactly the same f-value (corresponding to
this peak), even if they have different ion scores (see link 3).

In particular, the attached distribution was obtained from a search
with a concatenated decoy database containing reversed hits.
In this case I could decide whether to use the semi-supervised
approach or the non-parametric approach.
As you can see, with the non parametric approach, the odd peak is
fully associated with the positive model and I can get a higher number
of identified peptides.

I have no idea about the nature of this sharp peak, but it scares me a
bit, because I do not know whether I can trust my results.
Is it an artifact of the algorithm to obtain discriminant scores from
Mascot scores?

Would you trust these results?

(Nested Question: How shall I decide between semi-supervised and non-
parametric approach?)


We have been facing this issue for a while now, so I hope that someone
can help me.

Best regards,
Salvo

Link 1:
https://docs.google.com/leaf?id=0BxCqzqYo07lLNjQ5ZjIwMTMtODc3ZC00YzY2LTkzYzAtNjU0NTk4NzFkNzc2&hl=en&authkey=CKWfjc4N

Link 2:
https://docs.google.com/leaf?id=0BxCqzqYo07lLZjhkYTgxZTctZDRjMS00NjM0LTk2YWUtZTQ3ZjE5M2I1YjE3&hl=en&authkey=CIG57PsO

Link 3:
https://docs.google.com/viewer?a=v&pid=explorer&chrome=true&srcid=0BxCqzqYo07lLNWI1YmQ0ZGUtMjllNS00ZDJmLWE5MTEtMGYzZWNlNDAyOGY4&hl=en&authkey=CNXIkdMI

-- 
You received this message because you are subscribed to the Google Groups 
"spctools-discuss" group.
To post to this group, send email to [email protected].
To unsubscribe from this group, send email to 
[email protected].
For more options, visit this group at 
http://groups.google.com/group/spctools-discuss?hl=en.

Reply via email to