• Sign In
  1. Requests

isotopes/labeled forms not reported in mProphet command line export

support

View Request

view list print
isotopes/labeled forms not reported in mProphet command line export Will Thompson  2024-01-30 11:44
 

We have been looking at utilizing the mProphet command line export to help fix incorrect candidate peak assignments, and we notice that none of the isotope labeled, iRT, or surrogate standards are in the report. Is this on purpose or a bug? This removes the ability to fix misassignments of some of the isotope labeled standards in our workflows, particularly for small molecules where some isotope labeled standards share a mass. Skyline document attached; specifically the mProphet entries are missing for molecules in the 'iMT' and "IS" molecule lists.

CHeers

Will
 
 
Nick Shulman responded:  2024-01-30 12:25
Yes, all of those sorts of standards are excluded from the mProphet peak scoring model.
When the mProphet peak scoring was implemented, it was assumed that standards would not have the correct sort of target versus decoy behavior that would be necessary for the algorithm to be able to determine feature weights.

We could probably change the behavior of the "File > Export > mProphet Features" menu item (and the matching commandline command) so that, even though these molecules do not participate in the model training, Skyline would still include their numbers in the output file.

By the way, you did not seem to attach anything to this support request. If the problem was that the files you were trying to attach were more than 50MB you can upload them here instead:
https://skyline.ms/files.url
-- Nick
 
Will Thompson responded:  2024-01-30 14:04
Uploaded document to support request. Annotated with "Thompson support request 63323"

Will
 
Will Thompson responded:  2024-01-30 14:39
Thanks Nick! Yes, I'd like to request the additional compounds be included in the report when using File/export/mProphet features so that we don't mix our peak-identification algo with yours, for the standards. Better I think to use one model for the list of candidate peaks. Possible in the near-ish term?

Cheers

Will