• Sign In
  1. Requests

Q-value in groupwise comparisons

support

View Request

view list print
Q-value in groupwise comparisons nreimers  2025-04-24 17:10
 

Hello,

I am hoping to get some more information on the application of the q-value filter when setting up groupwise comparisons on the protein level in Skyline version 24.1. I am preparing to run pathway analysis on proteins that have been identified from DIA proteomics data in Skyline, so I am trying to generate high-quality groupwise comparisons with a p-value and log2 fold change for each protein. Implementing a q-value filter of 0.01 results in many missing abundances in my comparisons.

  • From my understanding, the q-value essentially descibes the FDR for picked peaks. If it applies to individual peaks (ie, peptides), how is it calculated on the protein level?

  • If I've already filtered my data by removing duplicate epptides, removing missing peaks, decoy scoring, etc. earlier in the analysis process, is it necessary to apply the q-value filter? Why is it applied in the groupwise comparison step and not earlier in the analysis process?

  • Would you recommend applying the filter for this application?

Thank you for your help!
 
 
Nick Shulman responded:  2025-04-24 18:11
The Q-value cutoff in the "Edit Group Comparison" window is intended to be used with the "Use zero for missing peaks".

Skyline has the problem that it often picks the wrong peak when the analyte is below the limit of detection. So, if you are doing a group comparison and one of the cohorts has an undetectable amount of protein the number reported by Skyline is often essentially a random number instead of the low number that would be more accurate.

The Q-value that the dialog is referring to is the one that is calculated by the trained peak scoring model, and is a number representing the probability of the peak being incorrect for that peptide in that particular replicate.
I believe it was shown in the following paper from 2015 that using the q-value cutoff combined with using zero for missing peaks resulted in good group comparisons:
https://pubs.acs.org/doi/full/10.1021/acs.jproteome.6b00881

We hope to come up with a better solution soon for quantifying undetectable peptides in Skyline.
-- Nick
 
nreimers responded:  2025-04-29 15:55
Thank you for the response. This makes sense for peptides, but how does it apply to proteins? I have observed scenarios when all the replicates in a group have a relatively high abundance for a given protein except for 1 which will have a 0 when I apply the filter. Would it use a 0 for a protein abundance when only one pepide peak can't be found?
 
Nick Shulman responded:  2025-04-29 19:30
When you are calculating fold changes at the protein level, if a particular peptide in a particular replicate has a q-value below the cutoff, and if "use zero for missing" is checked then it will be as if the transition peak areas for that particular peptide are all zero.

At the protein level, just like at the peptide level, a group comparison is performed by dividing one sum of transition areas by another sum of transition areas.
--Nick