I would just like to second what Hannes says: when using PeptideProphet,
never suppress the output of any negative results from the search engine
(X!Tandem or others) no matter how bad. PeptideProphet uses all this
information to model and discriminate between the correct and incorrect
identification distributions.

While I'm on the soap box, I'd like to mention another "don't": never
suppress from your sequence fasta file any sequences that you know or
strongly suspect are in your sample. If you know you have E coli
contamination that you don't care about, do not leave out the E coli
sequences. Do include all the discoverable sequences, let the search engine
find them, let the prophets model them, and then discard those IDs if you
don't care about them.

Eric


-----Original Message-----
From: [email protected]
[mailto:[email protected]] On Behalf Of Hannes Röst
Sent: Friday, April 04, 2014 1:27 AM
To: [email protected]
Subject: Re: [spctools-discuss] Can PeptideProphet and iProphet succeed with
few matched decoys?

Dear Vadim

I just checked my xtandem ini files and we have used these parameters

  <note type="input" label="output, results">all</note>^M
    <note>values = all|valid|stochastic</note>^M
  <note type="input" label="output, maximum valid expectation
value">0.1</note>^M
    <note>value is used in the valid|stochastic setting of output,
results</note>^M

according to the documentation, this will write out all results without any
E value filtering. As far as I understand the PeptideProphet algorithm, it
is necessary to report all results if you want to use PeptideProphet and
very likely the modelling will not work as expected when you perform any E
value filtering. I would thus suggest to also set the output, results value
to "all". I have attached the xml files that I currently use for your
convenience.

If you are interested in how we created the spectral libraries for the
OpenSWATH paper, please consult the method section available in the online
supplementary. There we describe our searches and how we converted the
X!Tandem searches to SpectraST and then TraML files.

I hope that helps

Hannes


On 4 April 2014 01:24, Vadim Patsalo <[email protected]> wrote:
> Dear Hannes, thank you for your reply.
>
> By "filter," I assume you mean the "maximum valid expectation value" in
> the output and refine settings of X!Tandem?
> If so, I've performed the following experiment.
>
> Evalue 1e-2:    8 decoys / 24280 non-decoys
> Evalue 1e-1:  105 decoys / 31902 non-decoys
> Evalue 1e0:   793 decoys / 38363 non-decoys
> Evalue 1e1:  3197 decoys / 42706 non-decoys Evalue 1e2:  4642 decoys /
> 44256 non-decoys
>
> Thank you for your advice -- I will attempt to build the Pos and Neg
> distributions using the more liberal cutoffs.
> My goal is to obtain a SpectraST library to interrogate SWATH datasets,
> using OpenSWATH, of course!
>
> What kind of decoy to non-decoy ratio is satisfactory for FDR modelling,
> in your experience?
>
> Vadim
>
> On Apr 3, 2014, at 11:28 AM, Hannes Röst <[email protected]> wrote:
>
>> Hi Vadim
>>
>> This will most likely _not_ work. It would probably be better if you
>> do not filter before xinteract but give it the full X!Tandem output
>> and then filter afterwards based on the computed probabilities. This
>> way you also might increase the total number of retained hits at a
>> fixed FDR.
>>
>> Hannes
>
> --
> You received this message because you are subscribed to the Google Groups
> "spctools-discuss" group.
> To unsubscribe from this group and stop receiving emails from it, send an
> email to [email protected].
> To post to this group, send email to [email protected].
> Visit this group at http://groups.google.com/group/spctools-discuss.
> For more options, visit https://groups.google.com/d/optout.

--
You received this message because you are subscribed to the Google Groups
"spctools-discuss" group.
To unsubscribe from this group and stop receiving emails from it, send an
email to [email protected].
To post to this group, send email to [email protected].
Visit this group at http://groups.google.com/group/spctools-discuss.
For more options, visit https://groups.google.com/d/optout.

-- 
You received this message because you are subscribed to the Google Groups 
"spctools-discuss" group.
To unsubscribe from this group and stop receiving emails from it, send an email 
to [email protected].
To post to this group, send email to [email protected].
Visit this group at http://groups.google.com/group/spctools-discuss.
For more options, visit https://groups.google.com/d/optout.

Reply via email to