Document Grid: Molecules missing Predicted Retention Time and/or Average Measured Retention Time when using a Spectral library

support
Document Grid: Molecules missing Predicted Retention Time and/or Average Measured Retention Time when using a Spectral library jtsorren  2022-12-13 14:41
 

Hi,

As the titled states I am using skyline-daily in molecule mode. When I add all of my molecules from a spectral library from the Spectral library Explorer I find that if I navigate to View -> Document Grid -> Reports -> Molecules the columns with either Predicted Retention Time or Average Measured Retention Time contain #N/A.

I am wondering if one of the columns should have picked up the molecule retention time given that the spectral library contains scans at explicit retention times. Maybe there is a setting I am not correctly using...

Also I am wondering if this is the cause of some incorrectly selected peaks for precursor that fall outside of the observed retention time seen in my spectral library for some molecules.

Thank you!

 
 
Nick Shulman responded:  2022-12-13 16:14

Can you send us your Skyline document?
In Skyline you can use the menu item:
File > Share
to create a .zip file containing your Skyline document and supporting files including spectral libraries.

If that .zip file is less than 50MB you can attach it to this support request.
You can upload larger files here:
https://skyline.ms/files.url

-- Nick

 
jtsorren responded:  2022-12-13 18:38

Here it is. Thank you

 
Nick Shulman responded:  2022-12-14 13:56
There is no way to see the library retention times in the Document Grid.
We have talked about adding this as a feature for a long time, but we have not done this yet.
One way to see the retention times would be to go to:
View > Spectral Libraries

You can also see the library retention times when you are looking at a chromatogram graph by right clicking on the chromatogram graph and choosing something from the "Peptide ID Times" submenu.

What were you hoping to do with the library retention times in the Document Grid? Would you want to be able to see all of the ID times numbers, or the average number, or something else?
-- Nick
 
jtsorren responded:  2022-12-14 20:33
Yes, I would like to view the average retention times for each molecule in the library.

More importantly, I noticed when I import DIA results occasionally peaks are picked that do not correspond to the retention time of the molecule in the library. Is there some settings I should change to make sure that doesn’t happen. If not, I can always manually shift the peaks…
 
Nick Shulman responded:  2022-12-14 21:00
There is a relatively new feature:
View > Other Grids > Candidate Peaks
which will show you the peaks that Skyline detected and how they scored.
If a peak overlaps with an identification in the library, then that greatly increases the peak's score, and usually results in Skyline choosing that peak.

If you set the "Explicit Retention Time" and "Explicit Retention Time Window" on a molecule, then that will constrain where Skyline looks for peaks.

If you send us a document where Skyline is badly picking peaks, we might be able to give you a better explanation of why Skyline is doing what it is doing.
-- Nick
 
jtsorren responded:  2022-12-19 13:04
Thank you! I will set the Explicit Retention Time correctly.

Additionally, I have noticed that I cannot import a decoy list for molecule mode on skyline-daily. I have tried to use File -> Import -> Transition List... but when I switch to Molecules there is no Decoy column to select. Are decoys in molecule mode unsupported?
 
Nick Shulman responded:  2022-12-19 16:37
No, small molecule decoys are not supported by Skyline.

We always knew that Skyline would not be able to automatically generate decoys for small molecule targets. I am not sure we ever thought about whether if you gave us a list of decoy transitions could Skyline train a peak picking model.

Do you have a Skyline document with target and (what are supposed to be) decoy molecules in it? Can you send us that document and maybe a couple of raw files? We might be able to support this scenario in a future version of Skyline.
-- Nick
 
jtsorren responded:  2022-12-20 08:41
Hm, yes I would just like to supply a list of mass shifted decoys for a peak picking model and qvalue generation. From my thinking this should fit in line with the peptide workflow.

I can try and draw up an example document, but I am curious as to what I am not understanding that makes it significantly different from peptide mass shifted decoys?
 
Nick Shulman responded:  2022-12-20 14:36
Deciding on appropriate decoy molecules is a very complicated topic.
There are patterns to the masses of the transitions which will be found in your measured results, because all molecules are made of the same atoms.
If you just chose random numbers for your decoy transitions, then the peaks for your decoy molecules will end up getting much worse scores than your target molecules, because your target molecules follow those patterns (e.g. they have a water loss) but the decoy molecules do not.
Calculating the false discovery sort of involves seeing how many decoy molecules had peaks which scored better than your targets, but if the decoys were chosen badly, then that ends up not being a fair comparison. That is, Skyline might end up telling you that your target molecules are very likely to be there, but, in reality, the only conclusion you are really justified in making is that there are molecules in the sample which are made of the same atoms or groups of atoms as your targets.
(There might be someone on this support board who can give you a more scientific explanation).

There have probably been some scientific papers written about how to choose decoys for specific types of molecules, and that technique is probably very different depending on the type of molecule (lipids, glycans, etc.)
-- Nick
 
jtsorren responded:  2022-12-21 09:48
Thank you for the detailed explanation. I appreciate the consideration of proper decoy transition generation with regards to properly calculating the false discovery rate.

I am wondering if it will be in consideration to enable the decoy column when importing a transition list in molecule mode. That way, after proper consideration of decoy generation, I, as the user, can upload a list of decoy molecules and take full advantage of the advanced peak picking implementation in skyline.
 
Nick Shulman responded:  2022-12-21 12:15
There are other features that we could implement which would probably be closer to what you need.
One feature that we have discussed would be to allow you to tell Skyline what weightings to use on the feature scores in the peak picking model.
Right now, the way to accomplish that would be to use a text editor to edit the .sky file and change the numbers that are in the text there.
If you send me your Skyline document I could get you started on that.

The difficult thing about allowing you to specify that a transition list contains decoys is that decoy peptides have some more information than just whether or not it is a decoy. In order to assign a decoy peak a reasonable value for the "library dot product" score, Skyline has to do something special because the decoy peptide will not be able to be found in any spectral library. For this reason, Skyline remembers which peptide was used to generate the decoy peptide, and when Skyline needs a "library dot product" value, Skyline compares the decoy's peak areas to the library spectrum of the original peptide.

Another thing is that in order to effectively train a peak scoring model, you would probably need to have a lot (hundreds or thousands) of molecules in your document.

Anyway, if you send us your Skyline document we might be able to figure out how to make Skyline behave better for you.
-- Nick