Tide-search Percolator example on crux website

[Abhishek Dubey]

I notice that in the example of using tide and percolator for simple search on crux website, the percolator outputs:

Tossing out "redundant" PSMs keeping only the best scoring PSM for each unique peptide.
INFO: Calculating q values.
INFO: Final list yields 0 target peptides with q<0.1.

Does this mean that percolator did not find any significant target peptides ?

Also, I notice that Percolator by-default do not run target-decoy competition for separate target and decoy search. But in example on website, --post-processing-tdc is used.

I have a data output of crux which yields me no significant target peptides when target-decoy competition is used. However, if I use default mix-max method, I get significant target peptides. One more observation is that in target-decoy competition mode, even if I keep --testfdr 1.0, there are still no decoy in the decoy psms output file.

[William S Noble]

On Thu, Feb 20, 2020 at 1:36 AM Abhishek Dubey <abhishekdub...@gmail.com> wrote:

I notice that in the example of using tide and percolator for simple search on crux website, the percolator outputs:


Tossing out "redundant" PSMs keeping only the best scoring PSM for each unique peptide.
INFO: Calculating q values.
INFO: Final list yields 0 target peptides with q<0.1.


Does this mean that percolator did not find any significant target peptides ?

Yes, I believe that is correct. I am cc'ing the Percolator team so they can comment on the precise semantics of this warning.

 

Also, I notice that Percolator by-default do not run target-decoy competition for separate target and decoy search. But in example on website, --post-processing-tdc is used.

I am not sure what example you are referring to.  In general, if you want to use TDC you need to either run a concatenated target/decoy search or you need to enable post-processing-tdc.  Otherwise, with a separate target/decoy search you can use the mix-max estimator.

 

I have a data output of crux which yields me no significant target peptides when target-decoy competition is used. However, if I use default mix-max method, I get significant target peptides.

As long as you are using mix-max with separate targets and decoys, this should be fine. In general, mix-max offers slightly better statistical power than TDC.

 

One more observation is that in target-decoy competition mode, even if I keep --testfdr 1.0, there are still no decoy in the decoy psms output file.

I will let the Percolator folks comment on this one.

Bill

 

--
You received this message because you are subscribed to the Google Groups "crux-users" group.
To unsubscribe from this group and stop receiving emails from it, send an email to crux-users+...@googlegroups.com.
To view this discussion on the web visit https://groups.google.com/d/msgid/crux-users/17167a00-37aa-44a7-9bd5-ce564b338dba%40googlegroups.com.

[Matthew The]

On Monday, February 24, 2020 at 5:36:04 AM UTC+1, Bill Noble wrote:

On Thu, Feb 20, 2020 at 1:36 AM Abhishek Dubey <abhishekdub...@gmail.com> wrote:

I notice that in the example of using tide and percolator for simple search on crux website, the percolator outputs:


Tossing out "redundant" PSMs keeping only the best scoring PSM for each unique peptide.
INFO: Calculating q values.
INFO: Final list yields 0 target peptides with q<0.1.


Does this mean that percolator did not find any significant target peptides ?

Yes, I believe that is correct. I am cc'ing the Percolator team so they can comment on the precise semantics of this warning.

Yes, that is correct.

 

 

Also, I notice that Percolator by-default do not run target-decoy competition for separate target and decoy search. But in example on website, --post-processing-tdc is used.

I am not sure what example you are referring to.  In general, if you want to use TDC you need to either run a concatenated target/decoy search or you need to enable post-processing-tdc.  Otherwise, with a separate target/decoy search you can use the mix-max estimator.

 

 

If neither of --post-processing-<x> flags is set, Percolator tries to detect what the type of input is. If we detect separate searches (i.e. we see multiple PSMs per scan+expMass combination) we effectively set the --post-processing-mix-max flag, if we detect a combined search we effectively set the --post-processing-tdc flag, you should be able to see this somewhere in the log file.

 

I have a data output of crux which yields me no significant target peptides when target-decoy competition is used. However, if I use default mix-max method, I get significant target peptides.

As long as you are using mix-max with separate targets and decoys, this should be fine. In general, mix-max offers slightly better statistical power than TDC.

 

One more observation is that in target-decoy competition mode, even if I keep --testfdr 1.0, there are still no decoy in the decoy psms output file.

I will let the Percolator folks comment on this one.

In target-decoy competition mode, only one PSM is retained per scan+expMass combination. If each of the targets scores higher than their decoy counterparts the decoy psms output file will be empty. However, in that case it is strange that you get no significant hits.. Perhaps it would be helpful if you could post your percolator log file.

-Matthew

 

Bill

 

--
You received this message because you are subscribed to the Google Groups "crux-users" group.

To unsubscribe from this group and stop receiving emails from it, send an email to crux-...@googlegroups.com.

[Abhishek Dubey]

I have uploaded the stderr, target psms and decoy psms for both mixmax and tdc : 

tdc stderr, target psms, decoy psms

mixmax stderr, target psms, decoy psms

 

On Monday, February 24, 2020 at 6:11:21 PM UTC+5:30, Matthew The wrote:

On Monday, February 24, 2020 at 5:36:04 AM UTC+1, Bill Noble wrote:

On Thu, Feb 20, 2020 at 1:36 AM Abhishek Dubey <abhishekdub...@gmail.com> wrote:

I notice that in the example of using tide and percolator for simple search on crux website, the percolator outputs:


Tossing out "redundant" PSMs keeping only the best scoring PSM for each unique peptide.
INFO: Calculating q values.
INFO: Final list yields 0 target peptides with q<0.1.


Does this mean that percolator did not find any significant target peptides ?

Yes, I believe that is correct. I am cc'ing the Percolator team so they can comment on the precise semantics of this warning.

Yes, that is correct.

 

 

Also, I notice that Percolator by-default do not run target-decoy competition for separate target and decoy search. But in example on website, --post-processing-tdc is used.

I am not sure what example you are referring to.  In general, if you want to use TDC you need to either run a concatenated target/decoy search or you need to enable post-processing-tdc.  Otherwise, with a separate target/decoy search you can use the mix-max estimator.

 

http://crux.ms/tutorials/search.html 

This is the link to the example I was talking about. The percolator outputs : Final list yields 0 target peptides with q<0.1

However, target.psms.txt contains psms with q value 0 and the text which follows says "In the list above, all of the matches have q-values of 0, meaning that they are highly significant."

[Abhishek Dubey]

On Monday, February 24, 2020 at 6:21:06 PM UTC+5:30, Abhishek Dubey wrote:

I have uploaded the stderr, target psms and decoy psms for both mixmax and tdc : 

tdc stderr, target psms, decoy psms

mixmax stderr, target psms, decoy psms

In tdc, I observed that even if I set --test-fdr 1.0, still I get no decoy psms while the q-value of target psms remain same.