I believe that my colleague, Luis Mendoza has implemented this on the
models page in TPP 5.0.0 that is still pre-release.

-David


On Thu, Jun 30, 2016 at 2:07 PM, Jesse <[email protected]> wrote:

> David,
>
> Would it be too much to ask that those nice plots you replied with be
> incorporated in the additional analysis info page where the FDR tables
> are?  I find that to be a very nice and quick way to assess my c-level
> setting.
>
> Best,
> Jesse
>
> On Monday, June 27, 2016 at 4:27:56 PM UTC-7, David Shteynberg wrote:
>>
>> Hello Jesse,
>>
>> As you noted, the CLEVEL used in your analysis is set to 5.  This means
>> that any PSM with an f-val not higher than 5 standard deviations from the
>> estimated mean f-value of the false PSMs, will automatically be forced to a
>> probability of 0 during modeling.  This excludes all the high scoring
>> DECOYS from further analysis since they are given a probability of 0.  I
>> would suggest you lower your CLEVEL value to achieve more PSMs with lower
>> f-values but probabilities greater than 0 after PeptideProphet.  It appears
>> that the software is working as designed in this case. Normally you should
>> set your CLEVEL as low as possible to give decent modeling without throwing
>> away too many good PSMs.  You should be able to get decent results with a
>> much smaller CLEVEL on this dataset 0, 0.5 should all work well here, see
>> attached.  Also you can boost the correct peptide IDs if you enable the
>> iProphet modeling and analysis on this data.
>>
>> Cheers,
>> -David
>>
>> On Mon, Jun 27, 2016 at 3:52 PM, Jesse <[email protected]> wrote:
>>
>>> Also here are links to the pepxml files before and after peptide prophet:
>>>
>>>
>>> https://drive.google.com/file/d/0B4N83JlasjgsemxITVZUNGpWdVk/view?usp=sharing
>>>
>>>
>>> https://drive.google.com/file/d/0B4N83JlasjgsUE9iRmNEZU5zSEk/view?usp=sharing
>>>
>>>
>>> On Monday, June 27, 2016 at 12:27:50 PM UTC-7, Jesse wrote:
>>>>
>>>> Hello TPP team,
>>>>
>>>> I've running some files through PeptideProphet using the xinteract
>>>> interface, and I think I'm not getting enough decoy hits resulting in way
>>>> too many correct IDs reported (evidenced by manual spectra review of
>>>> hits).  Please see the attached model results where no setting of min.
>>>> Probability results in a 1% FDR and causes all results to be accepted.
>>>>
>>>> I think this is somehow related to the c-level parameter?  I tried
>>>> setting it even higher and that does decrease the IDs but there is always
>>>> the same number of true hits (second attached screenshot).
>>>>
>>>> Otherwise, maybe I need to search my data with looser tolerances to get
>>>> more decoy hits?
>>>>
>>>> Your help is appreciated.
>>>>
>>>> Best regards,
>>>> Jesse
>>>>
>>> --
>>> 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 https://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 https://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 https://groups.google.com/group/spctools-discuss.
For more options, visit https://groups.google.com/d/optout.

Reply via email to