multiple similarity scores

rformassspectrometry / MetaboAnnotation

High level functionality to support and simplify metabolomics data annotation.

https://rformassspectrometry.github.io/MetaboAnnotation/

15 stars 9 forks source link

multiple similarity scores #3

Open michaelwitting opened 3 years ago

michaelwitting commented 3 years ago

Tools like MS-Dial state the forward and reverse dot product as well as number of matching peaks / total number of library peaks. Thinking about a function such as matchSpectra, I'm wondering if we should do the same? This seems to be standard now. @jorainer Any thougths?

michaelwitting commented 3 years ago

What about a rematchSpectra function that takes a matchedSpectra and reprocesses everything or adds an additional similarity score based on a different function.?

jorainer commented 3 years ago

Hm, the problem with that is that the MatchedSpectra supports only a single n:m mapping - if you re-match spectra you will most likely get different matching spectra and you will end up with multiple n:m mappings. That will be tricky to handle and I am afraid that it might also get the user confused.

What about a MatchForwardReverseParam that matches the query based on the forward score (that's an outer join?) and then calculates and reports also the reverse score (that's a right join?) for the matching spectra and also the number of matching peaks / total number of library peaks? We could get thus everything in one call...

michaelwitting commented 3 years ago

Okay, let's go for MatchForwardReverseParam.

jorainer commented 3 years ago

Just to get it right then:

the forward match (on which target spectra are matched) is based on an outer join (not a left join), correct?
the reverse match (that we just use to report an additional score with the matched target spectrum from above) is based on a right join?

I would then hard-code these joins into the compareSpectra calls.

michaelwitting commented 3 years ago

Yes, the forward is an outer join and reverse a right join assuming that the library spectrum is the "right" spectrum. Here are some figure explaining this: http://prime.psc.riken.jp/compms/msdial/download/mathematics/MS-DIAL%20FAQ-vs2.pdf (Slide 21)

jorainer commented 3 years ago

Thanks. OK, I will work on that.

michaelwitting commented 3 years ago

And what is left for me? ;-)

jorainer commented 3 years ago

adding documentation and unit tests for the matchMz (or annotateMz?) functions :)

michaelwitting commented 3 years ago

Let's call it better matchMz, fitting to the MS2 level.