Retention Times Score to Run Regression questions

Retention Times Score to Run Regression questions roman sakson  2022-04-22

Hi Skyline Team,

I am exploring the accuracy of various RT predictors wenn applied to my experimental data. Skyline is (as always) extremely useful here! Normally, I am using the residuals plot. I would like to understand the visualization a bit better and have some questions:

When applied within a document with several runs I can additionally get a score to run regression for all replicates or for the "best" one. I assume that these replicates have nothing to do with possible condition replicates annotated in my document and that Skyline just takes all present results as replicates here? Is the empirical retention time for each peptides just being averaged if I choose to plot all replicates or how are those combined? How is the "best" one chosen?

Thanks a lot,

Nick Shulman responded:  2022-04-22
For each peptide in the document, Skyline decides which replicate is the "best" replicate for that peptide.
Usually, "best" is decided by finding the replicate with the greatest total peak area. If you have trained an mProphet scoring model, then the "best" replicate is the one with the best Detection Z Score.
The code for deciding which is the best replicate is in the function "CalcBestResult" here:

If you have chosen "Replicates > All", then the number that gets plotted on the graph is the "Average Measured Retention Time", which is the average apex peak time across all the replicates. "Average Measured Retention Time" is a column which is available in the Document Grid.
-- Nick
roman sakson responded:  2022-04-22
Hi Nick,

thank you for making this clear now!

Just as an idea, I think that this plot is very useful to evaluate chromatographic conditions for bigger, scheduled datasets. Is there any chance to add some additional features to it, which are available for replicate comparisons plots? I am thinking about the possibility to highlight more than one target peptide at once (to visualize a set of iRTs in the middle of other peptides, for instance, see attachment). If all replicates have been selected, being able to automatically display error bars in y-axis dimension would be amazing! Of course, this is just my idea and I understand that you probably have a long list of more important To Dos, but maybe one day ... :).

Thanks again,