iRT definition

[hcji]

Hi,

Thanks for your great work, but I have a question: is the definition of the iRT predicted by prosit the same as the paper referred by the reference "Using iRT, a normalized retention time for more targeted measurement of peptides"? (doi: 10.1002/pmic.201100463)

Because I predicted the iRT of the iRT peptides with prosit and the pre-trained model, I got:
LGGNEQVTR | -9.96795
GAGSSEPVTGLDAK | 24.7797
VEATFGVDESNAK | 37.5635
YILAGVENSK | 48.0256
TPVISGGPYEYR | 53.7346
TPVITGAPYEYR | 59.9253
DGLDAASYYAPVR | 72.0672
ADVTPADFSEWSK | 82.2792
GTFIIDPGGVIR | 108.194
GTFIIDPAAVIR | 126.755
LFLQFGAQGSPFLK | 136.645

But in the reference is:
LGGNEQVTR | −24.92
GAGSSEPVTGLDAK | 0
VEATFGVDESNAK | 12.39
YILAGVENSK | 19.79
TPVISGGPYEYR | 28.71
TPVITGAPYEYR | 33.38
DGLDAASYYAPVR | 42.26
ADVTPADFSEWSK | 54.62
GTFIIDPGGVIR | 70.52
GTFIIDPAAVIR | 87.23
LFLQFGAQGSPFLK | 100

There is a significant difference.

[tkschmidt]

Hey hcji,
iRT is a concept of how to reference retention time within a given set of peptides.
Meaning: we took two peptides within the ProteomeTools project and defined them as 0 (peptide A) and 100 (peptide B) of our iRT scale. The rest of the data was then linearly scaled to those two reference points in all measurements and afterward used for training and prediction. Escher did something similar on their peptide set.

Now, if you want to apply it to some measured iRT or RT, you have to transfer the predicted iRT in your iRT/RT space.
Therefore, you can use simple linear regression and create a fitting function between iRT_Prosit and iRT of yours.

So by using 0,8485 * predicted_prosit_irt - 18,537 you can reasonably well create Escher iRT values.

Most commonly, slightly advanced methods like LOESS are used for creating this fitting function.

I hope this answer helps you, but you can find more information about the fitting of the iRT model in our method section.
but it is not meant as a fixed