iRT regression based failure to import DIA runs

iRT regression based failure to import DIA runs kguehrs  2020-07-09

Dear Skyline team,

I know that similar issues were already discussed but I am somewhat frustrated and not sure how ti further proceed.

I did 6 DIA runs for 2 preparation schemes of 3 samples similar to 3 biological replicates for 2 conditions. I added the Biognosys 1 peptides iRT reference to all sample and also used the Biognosys 11 calculator in Skyline in the peptide settings.

Four of the six runs were imported without problems. The remaining two run were not imported with the failure message shown in one of the screenshots in the attached zip archive. I added 6 additional figures that show the regression line for each of the samples together with the extracted peaks for one of the Biognosys 11 reference. We use for some special reason rather complex gradients that are optimized for the main samples that are often measured and that might be one possible explanation for the deviations of the retention times of some of the iRT peptides, but this behaviour is shared wth all samples.

I have tried to use different iRT calculated based on the Biognosys 11 with several samples peptides added but without success. I have also tried to use a Biognosys reference with only 9 or 10 peptides but this did not result in successful import. I also reduced the transitions to 3 product ions only without success of reimport.

I have asked a colleague to use the dataset with Biognosys software and he managed to import the sampled without any problem. Therefore, I assume that the problem is in Skyline and you might think one more time about the idea of r*r>=0.99 for the iRT regression being the most appropriate setting for import of DIA data that are more noisy and not that simple as PRM chromatograms.

Another suggestion relates to the import procedure. As far as I understood from previous discussions of the import issue the r*r>=0.99 for the iRT regression is calculated at the beginning of the import process but the error message is only shown after finalizing data import. In my trials I always had to wait some 10 minutes (rather old computer) until I had to realize the failure. If my assumption of the process timing is correct it should be possible to give an early message and possibility that the user can cancel the process to change settings and retry more early.

Perhaps, you can supply some more ideas how to progress as I am running out of ideas for further steps.

Thank you for your good work to offer Skyline for the community.

Brendan MacLean responded:  2020-07-09

Hi Karl-Heinz,
Sorry for all of the confusion. To me, it looks like all of your regressions shown for each replicate are >0.99. However, it looks like you have done some manual integration adjustment on the iRT peptides. Though, it is mostly moving the boundaries inward around the peak, which should still be the highest point and therefore the reported RT.

Would you mind posting a document with only the iRT peptides in it and the 2 problem runs imported. When the RT alignment fails, only the iRT peptides get imported, which should still allow you to assess what is going wrong with how Skyline is integrating those peptides, versus the expected iRT values. I am just not seeing the issue in the plots you sent. So, I am assuming it has something to do with the manual integration adjustments.

Also, something I find strange is that the replicate C3_Tr_DIA34_750ng is showing "ID" annotations. This should only happen when the data file you are importing is present in your spectral library. Since, I am assuming your spectral library came from DDA data, this would seem to imply that you gave your DDA data and DIA data the same name. Though, the "DIA34" in the name seems unlikely to have been used for DDA data. Can you explain the source of your spectral library? Did you use a tool like DIA-Umpire to search your DIA34 replicate and include it in your library with DDA data?

Your "Peptide IDs in Other Runs" (cyan lines) seem to indicate that there were other runs with a lot of IDs spread over the displayed gradient, which does seem to imply the presence of DDA data in the spectral library.

Thanks for your post to the Skyline support board and the screenshots. I am sure we can get to the bottom of this, and if we determine that it really should not be failing, I am sure we can get a fix into Skyline-daily relatively quickly.


kguehrs responded:  2020-07-13

Hello Brendan,

Please apologize me answering that late but I decided to play around with the settings you proposed first by myself to get an impression what is getting on.

The results are actually rather confusing to me. It gives already different results depending on the source of the iRT peptides you import into Skyline. I do often use a user specified database containing a protein concatenated from the individual iRT peptides for searching of the peptides in DDA runs used to generate the iRT library. Then, the iRT peptides are already in the results and I do not need to add them (that was at least the idea behind). When I use the fasta containing only this mentioned protein as a source for the iRT peptides (file- import-fasta) the identified peaks obtained by import of the MS runs are already different from the ones when the iRT are selected in the way Skyline use to make a new DIA import experiment, although the peptides are identical and partially (actually very badly) identified by Skyline.

The original library was created from DDA runs. Actually, we also did some PRM with the same samples and I modified the original library creating some more dedicated to special aims ones. In the file I attach I used only the original one. That might be a difference to the situation the screenshots were taken from. The cyan lines are indeed from the DDA runs and are indeed not in a very narrow range. As I mentioned we do try to find an appropriate way to prepare samples without loosing or "biasing" the information we extract. Therefore, the samples are very differently composed and thus the retention times very in addition to the technical variability generated by the LC machine which is actually not that bad.

In one of the attached files I selected in the transition setting to import only y- and b-ions but obviously skyline did not care and also imported the precursors, which might be appropriate for DIA data but anyhow somewhat misleading. With the default new DIA experiment method of Skyline and only the iRT peptides selected for import the import procedure did not fail although one of the iRT peptides was not correctly identified. It is at the right position in the measurement and I gave a 5 min window from the predicted time in the import method setting but another peptide was selected that is far from the predicted time.


kguehrs responded:  2020-07-13

Hi Brendan,

uploading the archive failed. You can download from our institute share using the following link.