Issue 880: SSL format changes for Feature Detection

Assigned To:Brian Pratt
Opened:2022-03-08 by Brian Pratt
Changed:2022-03-30 by Brendan MacLean
Related Issues
2022-03-08 Brian Pratt
Title»SSL format changes for Feature Detection
Assigned To»Brian Pratt
Notify»Brendan MacLean (test)
from Brendan's email 3/8/2022:

"On a thread with Mike, Dani, and Josh Eckels we are discussing Dani's use of Hardklor for unknown feature detection in MS1 scans and subsequent chromatogram extraction with Skyline.

It is seeming like having a (small molecule) .blib file with integration information in it is likely to be useful, and allowing that to be built from SSL also helpful.

The key missing things from SSL right now:
start and end integration times
the ability to have an entry without an associated spectrum
Regeneron has expressed interest in being able to see 'associated spectra' with found features. That is if in DDA a MS/MS was sampled from a feature, but not identified as a known peptide or compound, they would like to see Skyline associate that spectrum with the found feature. So, these SSL files may still contain spectrum associations, but that will also need to be optional, because there will be features with no MS/MS spectra sampled."

2022-03-08 Brendan MacLean
NotifyBrendan MacLean (test)»Brendan MacLean

2022-03-30 Brian Pratt
It seems we should probably revisit "Issue 689: File > Export > Spectral Library losing information from small molecule documents" in light of this - Kaylie was confused that Skyline skipped over precursors without fragments, and without chromatograms, and now that we're going to have BlibBuild pay attention to those we should probably be symmetrical in Skyline's output.

2022-03-30 Brendan MacLean
NotifyBrendan MacLean»Brendan MacLean;Nick Shulman
To do this, I think we will need to ask whether to include peak boundaries when we export a .blib file from Skyline. I think this is not often what people will want, since the boundaries will only apply when importing the same data again into Skyline, and then it will force the integration to match what it was when you did the export without doing any peak scoring.

Possible, of course, but may be relatively infrequent that anyone wants that.

2022-03-30 Brian Pratt
I hadn't even thought about peak boundaries, I was just thinking about precursors without fragments and precursors without chromatograms. Would we need to ask about those as well?

2022-03-30 Brendan MacLean
Yeah. I think there is really no point in storing just apex retention times for precursors without fragment transitions. I have no idea at all what we might store in a .blib from a Skyline document when there are no chromatograms, i.e. not even any idea of what the retention time might be. Then the entry in the .blib becomes just an indication of a target/precursor that may be of interest, without any of the information we typically associate with having a spectral library (namely 1. relative fragment ion abundance, 2. retention time, optionally including peak boundaries)

I don't see a compelling use case... yet. Let's stay focused on supporting feature rinding by tweaking SSL and proving a complete feature finding to Skyline document workflow. That is likely to take a while without adapting File > Export > Spectral Library to export .blib files that look like feature finding .blib files.

2022-03-30 Brian Pratt
OK. PR #2027 awaits review for the feature finding (SSL files) -> BlibBuild -> Skyline aspects of this.

2022-03-30 Brendan MacLean
You can give me a demo tomorrow (Thursday). What I am not seeing is a test that actually imports raw data (or a subsetted mzML) proving the feature finding can be translated into integrated peaks. Have you done that yet manually with data provided by Dani?