When to use "Decoy hits to pin down negative distribution" in PeptideProphet

30 views
Skip to first unread message

Soroush F

unread,
Jan 18, 2021, 7:02:29 PMJan 18
to spctools-discuss
Dear all,

I was wondering when one should use "Use decoy hits to pin down the negative distribution" in the Peptide Prophet (run through Petunia)?

If I want to use Target-Decoy approach to estimate FDR in my analysis, should I use this box, or it has nothing to do with it?

Thanks,
Soroush

David Shteynberg

unread,
Jan 19, 2021, 2:59:56 PMJan 19
to spctools-discuss
Dear Soroush,

This option makes use of the TPP's semi-supervised learning mode, which can be used in conjunction with semi-parametric mode, at the user's discretion.  Until now, some search engines supported by TPP analysis could only be modeled by PeptideProphet using semi-parametric and semi-supervised that required decoys to be used in the search.  Going forward with the new version of the TPP (pre-release version 6.0.0) this is not going to be the case as we will be providing both types of models for more search engine results.  I hope this answers your question.

Cheers,
-David

--
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 spctools-discu...@googlegroups.com.
To view this discussion on the web visit https://groups.google.com/d/msgid/spctools-discuss/4592e133-201d-4893-9573-ddee56f1071dn%40googlegroups.com.

Soroush F

unread,
Jan 20, 2021, 11:44:55 PMJan 20
to spctools-discuss
Dear David,

Thanks so much for your time answering my questions.

I gather from your answer that this option is for TPP. If that is the case, why does the PeptideProphet probabilities changes when selecting the option? Am I missing something?

Many thanks.
Soroush

David Shteynberg

unread,
Jan 20, 2021, 11:54:12 PMJan 20
to spctools-discuss
This option changes the way mixture models in the TPP (PeptideProphet) are generated.  The probabilities are estimated based on the mixture models learned by PeptideProphet.  When the mixture models change so will the probabilities.

Soroush F

unread,
Jan 21, 2021, 10:27:34 AMJan 21
to spctools-discuss
Thanks so much David for your response. It makes sense.

All the best,
Soroush
Reply all
Reply to author
Forward
0 new messages