Skyline for small molecules - development plans? takahashi18423  2019-07-29
 

Hello. I am a user of Skyline for proteomics applications and have been interested in where the development of the software for small molecules is headed. There are several approaches to filtering/processing HRMS data for identifying small molecule drug metabolites that will be great to have a vendor-agnostic data analysis solution for. In particular, algorithms such as background subtraction, isotope pattern filtering, or mass defect filtering would be key. Are any of these on the books for Skyline's development? I know it may be a long shot but I figured it is worth reaching out to find out.

 
 
Mike MacCoss responded:  2019-07-29

Thanks for your interest. We do currently do background subtraction from the chromatographic XIC peak area. Here is a short description of how that is done, https://skyline.ms/wiki/home/software/Skyline/page.view?name=tip_peak_calc

We also score analytes on how well they match a predicted isotope distribution. This score is called the idotp.

Is there something else that you were thinking of?

Mike

 
takahashi18423 responded:  2019-07-30

Hi Mike,

Thanks for the quick response. It looks like similar math but somewhat different problem. I am interested in filters that could be applied across entire data files to differentiate drug (analyte)-related vs matrix (or background) ions. For background subtract, this would be the full chromatographic run for test article minus the full chromatographic run for the matrix/blank. At each retention time, the full scan for the matrix sample is subtracted from the corresponding scan for the test sample. Theoretically only drug and drug-related species remain in the resultant file. The isotope pattern filter and mass defect filter would apply to the test sample file only and would filter out any ions that didn't meet the criteria of the expected isotope filter or mass change and mass defect around the test article, respectively. The filters and their application are described here: PMID: 21632546.

The processing is a major hurdle to making these algorithms usable but they are powerful tools for probing HRMS data files.

Ryan

 
Brian Pratt responded:  2019-07-30

Hi Ryan,

You'd have to really believe in your chromotography to do that kind of subtraction, though I'm given to understand that stable chromotography isn't necessarily a problem for small molecule work. Or is this a direct injection thing, with no chromatography at all?

Thanks

Brian

 
takahashi18423 responded:  2019-07-30

Hey Brian,

For sure, for the background subtraction to work effectively, the chromatography needs to be really solid. This hasn't been too much of a hurdle, as you mentioned, for small molecules. In the subtraction, some tolerances need to be built in to average across several scans in a time window because there will be some jitter in the elution times. The chromatography becomes less a concern (for the post-acquisition processing) for the isotope pattern or mass defect filter since they are only applied to the test sample and are dependent on parent drug/analyte mass ion properties.

Thanks!

Ryan

 
Brian Pratt responded:  2019-07-30

So you'd want to associate multiple blanks with a sample, and use the average of those at any given time point as the background at that time point?

 
takahashi18423 responded:  2019-07-30

There's never been a tool available to me that allowed multiple blanks (or samples) - it's been a single sample - single matrix. In that blank sample, a set of scans over a brief region of the chromatography were averaged and subtracted from the corresponding scans in the same region of the test sample. This removes some of the necessity for exact scan-for-scan matching between samples/chromatographic runs.

 
Mike MacCoss responded:  2019-07-31

Hi Ryan,
So it sounds like you are looking for a tool that looks for differences in signal between two groups of samples. Our lab used to do a lot of work in this area (see: https://pubs.acs.org/doi/10.1021/ac701649e). Basically, we would:

  1. Align all runs together in time.
  2. Bin data in time and m/z
  3. Look for bins that were different between conditions.
  4. Group bins based on chromatographic elution, isotope distributions, etc...
    We've definitely moved on to doing different things. If this is the strategy that you are interested in then XCMS might be an ok option.

That said, Pawel Sadowski gave a talk at our Skyline User Meeting (https://skyline.ms/wiki/home/software/Skyline/events/2019 User Group Meeting at ASMS/page.view?name=sadowski#) where he built the transition list targets by using a feature finding software tool first and then that was imported into Skyline. I have a student that is looking into building target transition lists using Hardklor (https://www.ncbi.nlm.nih.gov/pubmed/17580982) but that is a research project at the moment and not a development project.

I hope this helps.
-Mike