Re: [crux-users] Crux PSM search identify very few PSM with q<=0.01

[William S Noble]

Hi Sangeetha,

The Tide search engine does not automatically assign confidence estimates (i.e., q-values) to the PSMs it produces.  Not only does your search output not contain PSMs with q-values < 0.01; it does not contain q-values at all. You need to carry out an additional analysis step to do this, either by using the assign-confidence command (to simply apply target-decoy competition) or percolator (to use a machine learning approach).

Cheers.

Bill

On Fri, Sep 18, 2020 at 2:27 AM Sangeetha R <rsang...@cusat.ac.in> wrote:

Dear Professors,

I am a beginner in using the crux software. I have tried running the tide- search to extract all peptide spectrum matches. From the resulting xcorr score, when I calculate the number of PSM with a q-value threshold of 0.01, I am getting very few PSMs around 10. I doubted whether I have some mistakes in querying.  

the raw file used: Orbitrap XL CID-IT Yeast from https://chorusproject.org/pages/dashboard.html#/projects/all/819/experiments/1430/files

.raw file converted to mzxml file using msconvert tool (CID, MS2, CST peak picking parameters

Crux commands used:

crux tide-index --decoy-format shuffle --enzyme trypsin --max-length 21 --clip-nterm-methionine T --mods-spec C+57.02146 yeast.fasta yeast_ind 

crux tide-search --max-precursor-charge 3 --pepxml-output T --file-column T --top-match 1 --concat T --precursor-window 1 --precursor-window-type mass --remove-precursor-peak T B01a_130417_Yeast_long.mzXML yeast_ind 

Why I am getting very few PSM with q value threshold of 0.01.

--
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/44056b5e-e88c-47a3-b94e-a5d4d0a5b4dan%40googlegroups.com.

[Sangeetha R]

I have separately calculated the q-value for each PSM using xcorr score and target/decoy assignment in search result.

[William S Noble]

On Fri, Sep 18, 2020 at 9:57 AM Sangeetha R <rsang...@cusat.ac.in> wrote:

I have separately calculated the q-value for each PSM using xcorr score and target/decoy assignment in search result.

OK, then your results should agree with what you get by running assign-confidence on your Tide output file. Can you verify this and send me the log file?

Bill

 

To view this discussion on the web visit https://groups.google.com/d/msgid/crux-users/47ab3a49-f9d1-4084-a8e5-a8e31cb27538n%40googlegroups.com.

[Sangeetha R]

While running the assign-confidence on tide search result, the minimum q-value obtained is 0.08. Please see the attachment for the log file.

Query used:

crux assign-confidence --score "xcorr score" tide-search.txt 

[Sangeetha R]

Dear Team, 

Please help to identify the issue,  why I am getting zero PSMs with q-value threshold of 0.01 in tide-search?

I have tried using another yeast dataset given in 'http://noble.gs.washington.edu/proj/exact-pvalue'  as well. For this dataset also, I am getting zero PSM with a q- value threshold of 0.01. The query used is given below. The log files of tide-search and assign confidence is attached. 

crux tide-index --decoy-format shuffle --enzyme trypsin --max-length 21 yeast.fasta yeast_ind

crux tide-search --max-precursor-charge 3 --pepxml-output T --file-column T --top-match 1 --concat T --precursor-window 3 --precursor-window-type mass --remove-precursor-peak T yeast-01.ms2 yeast_ind --output-dir crux

crux assign-confidence --score "xcorr score" tide-search.txt --output-dir crux

 ( Parameters include fully tryptic peptides of maximum length 21, Precursor mass tolerance of +-3Da, maximum precursor charge +3, extracted top-scored peptide from target+decoy search. ) 

[William S Noble]

The problem appears to be that this is old data with low-resolution fragmentation spectra. Hence, rather than the default mz-bin-width of 0.02, you should use something like the 1.0005079:

--mz-bin-width <float> – Before calculation of the XCorr score, the m/z axes of the observed and theoretical spectra are discretized. This parameter specifies the size of each bin. The exact formula for computing the discretized m/z value is floor((x/mz-bin-width) + 1.0 - mz-bin-offset), where x is the observed m/z value. For low resolution ion trap ms/ms data 1.0005079 and for high resolution ms/ms 0.02 is recommended. Default = 0.02.

With this fix, I get 7962 PSMs at 1% FDR for the yeast data set.  The script I ran is attached, along with the log files.

Cheers.

Bill

To view this discussion on the web visit https://groups.google.com/d/msgid/crux-users/7a62c4e3-752c-44f1-93b0-6cad1f69daf1n%40googlegroups.com.

[Sangeetha R]

That really helped me. Thanks a lot for the response.

The issue got solved now.