ProteoformGroup

[acesnik]

I want to open this up as a discussion outside of the coming pull request. I'm excited for the coming functionality, and I don't want to slow that down.

Making "proteoform hypothesis" a more formal in the future makes me a little uneasy because a hypothesis generally leads to testable predictions based on former evidence. I can't seem to state that prediction:

• Question: "What is the localization of the modification X on proteoform Y?"
• Hypothesis: "The spectrum shows a fragment spanning amino acids 1-22 that has a mass difference characteristic of X. Therefore, the modification is somewhere along that sequence segment."
• Prediction: "If ... then ... "

It seems like the hypothesis should get to the bottom of the question: the exact localization. Then, the prediction would be something about the function or effect of that modification localization.

Instead, we're trying to represent a partial answer the the localization question, so what about one of these: PseudoProteoform, PartiallyLocalizedProteoform, PartialProteoform?

Does my concern make sense?

[rfellers]

First off, thanks for detaching this discussion from the forthcoming PR. I'm not married to "proteoform hypothesis," but we had to start with something!

Your concern absolutely makes sense. Here is my best attempt:

• Question: What proteoform (or proteoform space) explains our set of spectral observations?
• Hypothesis: We propose that this ProForma term (complete with any ambiguity) explains these observations.
• Prediction: If correct, then one should see a consistent precursor mass, multiple consistent fragment masses, etc.

So, in the case of our search, each proteoform in the database is a hypothesis that we test and score.
Does that make sense? I don't even know if it does :) Feedback would be great.

To me, PseudoProteoform implies that the proteoform is wrong or spurious and PartialProteoform implies that the proteoform is not complete or could be a subset of a proteoform. PartiallyLocalizedProteoform feels better, but still isn't satisfying because it may be completely localized. I feel bad criticizing because I know how hard this is. Bryan and I chatted for a couple hours and didn't come up with anything great.

[acesnik]

What about the ProteoformSpace term you used in the statement of the question? That could be fully localized or ambiguous, i.e. it could represent a single or a thousand proteoforms.

[rfellers]

Great job pulling that term out, never crossed my mind. That's not bad at all, I could go for that. What do others think?

[rmillikin]

I like ProteoformSpace to encompass the space of possible proteoforms given (in)complete data.

[leahvschaffer]

I like ProteoformSpace as well. Or maybe something like ProteoformCandidate(s)

[rleduc42]

Love it!

[acesnik]

One of the staff scientists here noted that ProteoformSpace almost sounds too large -- like it's spanning almost all proteoforms instead of just a few. ProteoformGroup might be a little more constrained, and a lonely group can also have only one member.

[b-early]

I think ProteoformSpace and ProteoformGroup both get the idea across well. Out of those two I prefer ProteoformGroup more, since I agree that it fits the scope better.

[leahvschaffer]

I like ProteoformGroup. One comment is that in bottom-up there are "protein groups" where the parsimony couldn't disambiguate the proteins in that group.

[rfellers]

After thinking about this a bit, I'm on-board with ProteoformGroup. I agree that "group" does seem to imply a more appropriate scope. I'll change the name in my next PR to reflect the change. This term would feature prominently in a ProForma v2.0 follow-up that adds support for ambiguity ...

@lschaffer2, when you mention the example from bottom-up, do you think that is a good thing or a bad thing?

[leahvschaffer]

I think this is a good thing - since the term is used for ambiguous items in a group in other areas of proteomics it makes it more intuitive/consistent.

[rfellers]

changed to ProteoformGroup in #43 ... but the conversation can continue. It is very easy to change again ...

[acesnik]

I still like ProteoformGroup ... I think it connects the concept to bottom-up "protein groups" nicely.