On the first page of the "Import Peptide Search" wizard, there is a place where you can specify the "cutoff score". If you set that number to a more lenient value than your other tools, then you might see some peptides that you did not see elsewhere.

When you import peptide search results into Skyline, I believe you will only end up with peptides that were found with at least some confidence in your results. If you are seeing peptides that really are not in your Proteome Discoverer output, then I cannot think of what could be going wrong.

It would be helpful if you could zip up all of your files and upload them here:
https://skyline.ms/files.url

If you want to train a scoring model in order for Skyline to use decoys and figure out which peptides are present, then you might want to look at the advanced peak picking model tutorial:
https://skyline.ms/wiki/home/software/Skyline/page.view?name=tutorial_peak_picking

However, that tutorial is mostly targeted at DIA, not MS1 DDA, which is what it sounds like you have.

Thanks Nick for prompt response. I will for sure upload files and let the ball roll.
Just to make sure that we are on the same page. We were using Sequest within Proteome Discoverer and please correct me if I am wrong but the "cutoff score" within "Import Peptide Search" wizard has a maximum value of 1 , which to me indicates possible TPP-Peptide/Protein Prophet scoring, which we did not have in this particular MSF file.
I am absolutely sure that those "ghost" peptide is not affiliated with any protein ( see enclosed) so its appearance in Skyline really surprised me. THose peptide was not present in the list of all ID peptides , nor you can find it in the list of all PSMs.
What might have been happening that Skyline was working extra-time in trying to fit available raw signal to possible signal from theoretical digest and happily reported it for that particular protein. May be this situation is not unique. Need to check other proteins. THough that is my wild guess. Will follow with MSF and FASTA. Indeed I am hardcore DDA so for now will refrain from DIA affiliated workflows.
Cheers
Victor

Hi Victor,
If you hover the mouse over the cut-off score control there is actually a tip that explains it. The direction of the value did originate with my familiarity with the TPP when I started Skyline, but it works for all scoring pipelines. You just have to use 1 - score for score types where 0 is the best score.

For example, if your pipeline produces q values and you want a 1% FDR, then you would use 1 - 0.01 = 0.99 as your cut-off.

Maybe we should add a checkbox to flip the sense of this field between 1 is best and 0 is best, which would just change which would just require us to actually deliver 1 - value to BiblioSpec when the user wanted to think of scores as 0 being best.

But, it works now, if you can think in terms of 1 - score.

My guess is that you have misunderstood this, ans used a cut-off like 0.01 = 1 - 0.01 = 0.99, which will allow for a 99% FDR when applied to a q value (even if you haven't achieved that).

The PSM you show certainly appears unlikely to have received a high score.

Please send a screenshot of the library details page (click the "..." button beside the library name) from View > Spectral Libraries for your library. This will contain the cut-off score used for each file. If that value is 0.01, then rebuild your library using 0.99 instead. In fact, if it is closer to zero than to 1, rebuild the library with 1 - value.

Sorry for the confusing implementation. Thanks for reporting it and helping use work through the details.

--Brendan

Hi Brendan,
Thanks for wading in. Hope you enjoyed your reef adventure. Indeed looks like my original cut-off of 0.95 was a bit lax and when I tightened things up to 0.99 all suspects disappeared and Skyline target window started to resemble PD output. The only question remained what would happen if I used cut-off value of 1 . Does it mean that only super-confident ID would remain ? I would not mind that.

Now that we solved this particular issue there are a couple of other smaller one. I mentioned that for some reason peptides and PSMs which were successfully quantified in PD were IDed but not quantified in Skyline even in the presence of good fragment signals and isotopic fidelity. This was especially true for peptides with multiple deamidation sites. In some cases the signal was quite prominent and I could see isotopes in Skyline window. But not XIC in Chromatogram. I am enclosing PPT illustrating this issue along with Peptide and Transition Parameter values.

Last thing. Sometimes when I am changing something in parameter Skyline target window is not automatically updated to reflect made changes. Sometimes it is happening and sometimes it is not. I did not get a hang on those things yet. Is there somewhere a button to force the program to update calculus taking into account made changes in the parameter settings ?

Cheers

Can I get that Transition Settings - Full Scan tab? That is one of the most interesting in this case, but was not included. Instead you duplicated the Instrument tab.

Thanks. Heron Island was quite nice. Lots of rays, sharks, turtles, birds. Looking forward to my next visit to Australia.

--Brendan

Enclosed. I played a bit with different settings there : changing RT filtering ( 2-4 min), Mass Accuracy (4-10 ppm) ,
Different Number of Isotope Peaks ( 1-5 ) or not at all. In all cases those changes did not have an effect on Quan of de-amidated species.

I have to think that you are expecting these settings changes to take effect immediately. When you make changes in the Transition Settings - Full-Scan tab or you add new targets, when you are working with chromatogram extraction from full spectra, you will need to re-import the files. You can use Edit > Manage Results - Re-Import (you must select the replicates you want to re-import).

Otherwise, I just can't see why precursor m/z values like 851 and 832 would not be extracted from your MS1 spectra.

It is not that Skyline is not finding anything interesting. When you don't see to colored (green or red) dots to the left of the text and icons in the Targets view. That means Skyline has no extracted chromatograms, which is pretty unusual for MS1 extraction if the targets were actually present during the results import.

If there were no signal at all for the m/z values over the retention time range of interest, then you would see red dots and a graph with the expected legend and some color along the x-axis.

It is also worth noting that Skyline does not change chromatogram information when you use Edit > Manage Results - Remove, unless you then Save. So if you just did Edit > Manage Results - Remove and then File > Import > Results of the file you just removed, then you would not see any changes.

I am sure there is some simple explanation of what is hanging you up on this. We just need to figure out what it is. Hope something I just said clicks. Let me know.

--Brendan

I was thinking that in general one of the potential complications in Quan of deamidated species is quite small difference in RT and possible overlap of isotopes. So to test Skyline capabilities to distinguish between them may be it is possible to reverse engineer Skyline parameters based on the attached XIC QUAN from XCalibur itself. I can see that RT difference between peaks is 1.5 min ( where minute is on 100 scale) and about 5.5 ppm difference in m/z. So what would be Skyline settings that would for sure differentiate based on RT alone taking into account only first isotope. I can see that deamidated species have a bit of contamination from first isotope of non-deamidated and that can complicate things a bit. In addition, I saw somewhere settings regarding ambiguity : may be this one could also play a role ? If ambiguity is not allowed that some of overlapping isotope signals already assigned for non-deamidated will not be reported for de-amidated. Something like that. Just my wild guess.
Also may be it is possible to separate those signals completely and process file with two different settings : First settings where de-amidation will be ignored, and second one where only demidated signals will be reported. A bit drastic but...

Please upload your Skyline document (using File > Share - Complete) and one of your raw data files to the location mentioned by Nick:

http://skyline.ms/files.url

I will have a closer look. It still doesn't really make any sense to me that Skyline would fail to extract chromatograms for so many targets as you have shown.

If you have removed all results, saved the document and then reimported your results and you still get that many targets without green or red dots on the left side, then we really need to look at the data to understand why.

--Brendan

Good news, Brendan. Re-importing really started things moving in the right direction.
As discussed I tightened RT and Mass Accuracy filtering and that
had a desired effect as now everything which should be quantified is quantified (enclosed).
If I understood correctly something which IDed ( 62.0 and 62.3) is going to be integrated, while peak at 60.6 min ( +3 ppm), that is visualized possibly due to an overlap with some isotopes of non-deamidated species, without ID is not going to be integrated.

Just a couple of other clarifying questions. Definition of RT and PPM filtering.
When I am setting RT filtering within 0.5 min I am setting it 0.5 min to the left and to the right of "expected" RT, correct ?
Same with PPM. THose 4 PPM mass accuracy is it 4 pmm both left and right from the peak or is it 2+2 ? I remember there was a difference in mass accuracy in different Skyline places.
I can still upload files if you wish to play with them a bit.
Cheers

I am glad you managed to re-import. Yes, your document looks much better with the colored (red, orange and green) dots beside all the targets as expected.

But, again you need to be careful on the Transition Settings - Full-Scan tab to distinguish between the settings as you have them at the moment and what they were when you imported your data. It seems pretty obvious from the chromatograms you are showing that you did not have your RT filtering at 0.5 minutes (yes +/-). It looks a lot more like +/- 3 minutes for RT filtering during extraction.

The filtering only impacts the chromatogram extraction. It does not impact peak integration after you have imported the data. So, you cannot extract +/- 3 minutes chromatograms and then set the filtering to +/- 0.5 and expect Skyline to suddenly confine its peak picking to just that 1-minute range despite the fact that it extracted 6 minutes.

But, yes, you have the meaning of the tolerances correct. They are +/- the number you give. So, +/- 3 minutes of RT filtering becomes 6 minutes of chromatogram. It is worth nothing that in the case of MS/MS IDs, the +/- 3 minutes will be around the entire range of IDs. So, if you have 10 IDs spanning 2 minutes, then your total extracted chromatogram range will be 8 minutes.

Hope this helps. Thanks for digging into the details and making a bit of progress.

--Brendan

Just to be clear on a theory side, Brendan.
If I am using Results of the Database search of a particular file to create a library to quantitate very same file I actually see no reason to be that generous with regards to RT filtering and probably should limit it to actual peak width ( give or take ) and thats about it. Probably in my particular case that would be plus minus 10 seconds, though tailing makes things a bit more complicated.

Mass accuracy can be also confined to maximum PPM value which could be easily extracted from PD. On mass accuracy note, probably there is no way to mass recalibrate in Skyline, and squeeze those mass accuracy parameter even tighter, right ?

With regards to digging, I am afraid Thermo did not leave me any other choice as somehow I feel PD is going backwards in their LFQ Quan module and we need to find some credible and verifiable platform for that. Next down the line will be creation of SSL libraries for LFQ of cross-links and disulphide bonds. Should be fun.
Cheers
Victor

With DDA data, frequently you will not see IDs in all runs. Skyline performs linear regression between the retention times of shared IDs between all runs. When an ID is missing in a run, it will use the linear equations from the regressions to transform the IDs in other runs into the RT space of the run without the ID, and then it will use these as the range around which to add your +/- retention time tolerance. This is the case with the most sensitivity to the tolerance you use.

Ideally, at least 1 of the aligned IDs would land inside the peak of interest, which would still mean your tolerance shouldn't need to be larger than 1 peak width. Though, in label-free quant, if you truncate the peak at all, you don't have a valid measurement. So, I might err a little on the side of caution, and maybe add a bit more than an expected peak width.

Skyline does not have a recalibration feature for mass accuracy. We are well aware of the concept, but have not implemented anything for it yet.

What platform are you using to identify your cross-linked peptides? Jimmy Eng and Yuval Boss have implemented an external tool for Skyline called Corss-Link Transition Calculator:

https://skyline.ms/skyts/home/software/Skyline/tools/details.view?name=Cross-link%20Transition%20Calculator

Which transforms Kojak results into Skyline general molecule documents:

http://www.kojak-ms.org/

Thanks for your effort to use Skyline in your research.

--Brendan

I am using up to 6 different software packages, including TPP-Kojak interface, and it is good to hear that one can transform Kojak results to data files acceptable by Skyline. Alternatively, recently there was a publication where custom build SSL library was used to quantitate results of XL experiments (enclosed). I think this approach could be actually applicable to both cross-linking and Disulphide-bond quantitation tasks. I requested example of this library but nothing materialized yet. I think one of the advantages of this SSL approach is its universality as to my limited understanding SSL file is just CSV file which need to have certain info, which is readily extractable from any XL program and thus making its universal tool. I need to clear a bit of items from deck and then we can open new thread and hammer this topic to conclusion for both DSB and XLs. May be create a tutorial cause quan of XL data is a bit of a grey area and I think Skyline is uniquely positioned to tackle it.
Cheers
Victor

I just returned from the Lorne Proteomics Symposium in Australia, and it was definitely my impression that interest in using Skyline for cross-linking studies is growing. Perhaps interest in using cross-linking in proteomics is itself growing.

Great to hear that you are interested in pushing the technology forward and including Skyline in your effort. We are happy to help as much as we can.

--Brendan