Adding Bonanza score to spectrum similarity calculation

[kozo2]

I will add bonanza.R to https://github.com/rformassspectrometry/MsCoreUtils/tree/main/R . I will mimic Spectral_Similarity_Algorithms.r in https://doi.org/10.1021/acs.analchem.0c02521 (for the implementation). https://github.com/rformassspectrometry/MsCoreUtils/blob/main/R/gnps.R also mimics Spectral_Similarity_Algorithms.r https://github.com/rformassspectrometry/MsCoreUtils/blob/main/R/gnps.R#L75-L78 . I have obtained permission from its author, @Philipbear , to mimic it.

[jorainer]

yes, you're right. we base the GNPS implementation on the same paper (Michael Witting got permission for that). Note that in MsCoreUtils/Spectra we split the spectra matching functionality into two parts:

• code to map/match peaks between the two spectra (based on ppm and tolerance).
• code to calculate the similarity based on the m/z and intensity values of the matched spectra.

So, for the first part you could use/rely on the join function. This is for example how compareSpectra in the Spectra works, we have parameter MAPFUN = joinPeaks and FUN = ndotproduct by default there, i.e. use join for the matching and then the normalized dot product for the similarity calculation (see also compareSpectra).

Alternatively, as in the case of the entropy-based similarity scoring, where we use the functionality from the external entropy package, it is also possible to perform both the peak matching and similarity scoring in the same functions (see here how that could be done). I would prefer the first approach, i.e. keeping the mapping and scoring separate.