Reintegrate -> Peak scoring error using mprophet

support
Reintegrate -> Peak scoring error using mprophet sas28  2018-06-08 07:07
 

Hello,
I was wondering if you could help with the following error:
System.IO.InvalidDataException: Insufficient target peaks (0 with 5718 decoys) detected at 15% FDR to continue training. ---> System.IO.InvalidDataException: Insufficient target peaks (0 with 5718 decoys) detected at 15% FDR to continue training.
at pwiz.Skyline.Model.Results.Scoring.MProphetPeakScoringModel.CalculateWeights(String documentPath, ScoredGroupPeaksSet targetTransitionGroups, ScoredGroupPeaksSet decoyTransitionGroups, Boolean includeSecondBest, Boolean nonParametricPValues, Double qValueCutoff, Double[] weights, Double& decoyMean, Double& decoyStdev, Boolean& colinearWarning) in c:\proj\pwiz_x64\pwiz_tools\Skyline\Model\Results\Scoring\MProphetScoringModel.cs:line 405
at pwiz.Skyline.Model.Results.Scoring.MProphetPeakScoringModel.<>c__DisplayClass7.<Train>b__4(MProphetPeakScoringModel im) in c:\proj\pwiz_x64\pwiz_tools\Skyline\Model\Results\Scoring\MProphetScoringModel.cs:line 255
at pwiz.Common.SystemUtil.Immutable.ChangeProp[TIm](TIm immutable, SetLambda1 set) in c:\proj\pwiz_x64\pwiz_tools\Shared\Common\SystemUtil\Immutable.cs:line 201 at pwiz.Skyline.Model.Results.Scoring.MProphetPeakScoringModel.Train(IList1 targetsIn, IList1 decoysIn, LinearModelParams initParameters, Nullable1 iterations, Boolean includeSecondBest, Boolean preTrain, IProgressMonitor progressMonitor, String documentPath) in c:\proj\pwiz_x64\pwiz_tools\Skyline\Model\Results\Scoring\MProphetScoringModel.cs:line 196
at pwiz.Skyline.SettingsUI.EditPeakScoringModelDlg.<>c__DisplayClass13.<TrainModel>b__b(IProgressMonitor progressMonitor) in c:\proj\pwiz_x64\pwiz_tools\Skyline\SettingsUI\EditPeakScoringModelDlg.cs:line 269
at pwiz.Skyline.Controls.LongWaitDlg.RunWork(Action1 performWork) in c:\proj\pwiz_x64\pwiz_tools\Skyline\Controls\LongWaitDlg.cs:line 228 --- End of inner exception stack trace --- at pwiz.Skyline.Util.Helpers.WrapAndThrowException(Exception x) in c:\proj\pwiz_x64\pwiz_tools\Skyline\Util\Util.cs:line 1842 at pwiz.Skyline.Controls.LongWaitDlg.PerformWork(Control parent, Int32 delayMillis, Action1 performWork) in c:\proj\pwiz_x64\pwiz_tools\Skyline\Controls\LongWaitDlg.cs:line 176
at pwiz.Skyline.Controls.LongWaitDlg.PerformWork(Control parent, Int32 delayMillis, Action`1 performWork) in c:\proj\pwiz_x64\pwiz_tools\Skyline\Controls\LongWaitDlg.cs:line 131
at pwiz.Skyline.SettingsUI.EditPeakScoringModelDlg.TrainModel(Boolean suppressWeights) in c:\proj\pwiz_x64\pwiz_tools\Skyline\SettingsUI\EditPeakScoringModelDlg.cs:line 272
at pwiz.Skyline.SettingsUI.EditPeakScoringModelDlg.TrainModelClick() in c:\proj\pwiz_x64\pwiz_tools\Skyline\SettingsUI\EditPeakScoringModelDlg.cs:line 191

The Skyline file contains 447 target proteins with the same number of decoys (shuffled sequence).
When performing -> Reintegrate -> Peak scoring, it gives error when mprophet option is used, default is fine but then Skyline integrated MSstats gives an error as it is also missing q-values:
** Reading the data for MSstats.....
** iRT proteins/peptides are removed.
Peptides, that are used in more than one proteins, are removed.
** Truncated peaks are replaced with NA.
** For DDA datasets, three isotopic peaks per feature and run are summed.
Error in SkylinetoMSstatsFormat(raw, filter_with_Qvalue = filter.qvalue) :
** Please check annotation for Condition and BioReplicat column. There is missing information.
Can't finish analysis.

What can i do?

Many thanks,
Silvia

 
 
Brendan MacLean responded:  2018-06-08 08:17

Hi Silvia,
This is essentially telling you that mProphet has failed to separate your targets from your decoys at the very outset and there are is insufficient separation to continue training. You will need to switch tabs to the individual scores tab and start looking at why they may be failing to separate targets from decoys. You should also look at which scores are enabled and in use.

Skyline will disable scores if they are not available for all precursors. For instance, the most valuable score in these analyses is usually the library dot-product, which may be disabled, if any of your targets have less than 3 fragment ions.

If you attach some screenshots of these views in the model training form, I may be able to help further.

--Brendan

 
sas28 responded:  2018-06-08 09:17

Indeed it looks odd. Any ideas why?

 
Brendan MacLean responded:  2018-06-08 15:12

Can you provide me the document (using File > Share - Complete to produce a .sky.zip) by uploading to the drop-point below?

http://skyline.ms/file.url

That will make it easiest for me to diagnose. Definitely, you are not going to be able to build a model from only the 3 available scores.

--Brendan

 
Brendan MacLean responded:  2018-06-08 17:45

I remembered when I saw something exactly like this (and it caused similar confusion), at the recent May Institute - Capstone Session. It turned out to be because the document had a settings error in the Peptide Settings - Modifications tab, where the "Internal standard type" had been set to "light".

It seems like this would explain what you are seeing as well, since in the mProphet pane, you are not even getting the intensity score of zero for everything (i.e. you have no "heavy" area), but your "Default" scores make more sense. This is likely because "Default" scores will use the standard areas when present. If you have "light" as your standard type, then your slides make sense.

On the other hand, we should probably detect this case and offer to correct it when you bring up the peak scoring model editor in a case like this. It is pretty obviously not what you want, and clearly very confusing.

--Brendan

 
sas28 responded:  2018-06-11 02:27

You were right. The peptide setting -> modification -> internal standard type was indeed set to light. Changing it to heavy sorted the problem. Thank you so much. Great help once again!

Silvia