problems with RT predictor christiee  2024-08-01 15:33
 

Dear Skyline Team,

I am hoping you can help me with a problem that is causing me great frustration.

Recently when loading LysC digested PRM data into skyline, there is no longer a predicted retention time associated with the data. This seems to be a relatively recent (we started noticing it around June) problem for us. I've attached two files to this message. The first is a file entitled HA_May2024. This data was acquired in may of this year and as you can see when you open the data, there is a predicted RT time associated with the data. The second file is entitled HA_V2d. This data was acquired on July 27th and as you will see, there is no predicted RT time associated with this data. Interestingly, if you remove the data from the May2024 file and import the data from the V2d file into the May2024 file, there is again no predicted RT time associated with the data. Perhaps even more perplexing is that if I take the data acquired in May and import it into the recently made V2d skyline file, the data has predicted RT times! To me this is suggesting that the HA_V2d skyline document (and associated blibs and iRT databases) are not the problem but perhaps there is a problem with the way the data is being read??? One thing to note is that we are only seeing this with LysC data, our Trypsin digested data does not have this problem.

If possible, could someone on your team take a look at these files and see if they can identify why this is happening.

Thank you for your time,
Christie

NOTE: the data for this request has been uploaded to your file sharing folder. The title of the file is problems_w_RT_predictor_20240801

 
 
Nick Shulman responded:  2024-08-01 18:07
Thank you for uploading that .zip file.

The reason that you are not getting predicted retention times in "HomeAlone_V2d.sky" is that when Skyline plots the iRT values of the iRT standards against their observed retention times, the line is not straight enough.

Skyline insists that the R value of the linear regression between iRT score and observed retention time be at least 0.99.
When you have 12 iRT standards, Skyline is willing to ignore up to 3 of them in order to achieve a linear regression with an R value above 0.99. (The number of standards that can be ignored is 20% of the total, rounded up).

You can see that Skyline was not able to do this if you go to:
View > Retention Times > Regression > Score to Run
Then, you have to choose the following from the right-click menu:
Points > Standards
Replicates > Single
and also choose "Set Threshold" and type in the number 0.99.

If you do that, you will see a graph like the one attached.
The light purple diamonds are the standard peptides which had to be ignored in order to achieve a linear regression with an R value above 0.99.
There are four light purple diamonds which is why you do not get a predicted retention time.

One way to fix this would be to recalibrate the iRT predictor.
Recalibrating means assigning new iRT values to all of the peptides based on where you actually observed them.

You would go to:
Settings > Peptide Settings
then, push the calculator button and choose "Edit Current".

In the "Edit iRT Calculator" dialog push "Recalibrate".
In the "Calibrate iRT Calculator" dialog push "Use Results".
Then, after you OK all of the dialogs the "Retention Times - Score to Run Regression" graph will become a perfectly straight line.

It is unfortunate that Skyline did not give you a better warning message in this scenario. I will try to figure out some way to make this work better in the next version of Skyline.
-- Nick