[Psidev-ms-dev] Handling variations in generation of m/z vs. intensity data arrays

[Wilfred H T. <Ta...@ap...>]

For a given mass spectrum, there are a large number of possible ways to 
generate m/z vs. intensity data arrays, depending on the answer to 
questions such as:
 - Entire m/z vs. intensity profile or peaks only?
 - All peaks or monoisotopic peaks only?
 - All peaks or "prominent" peaks only (must be above some signal-to-noise 
threshold, for example)?
 - "De-charge" peaks or not?
 - And so on...

I can think of 2 main ways to handle these variations (as well as options 
in-between that are a hybrid of these 2 extremes):
(1) The software that generates mzML asks the user in detail what his or 
her preferences are, and then the software generates the appropriate m/z 
vs. intensity data arrays.
(2) The software that generates mzML doesn't ask the user anything. Rather 
the software writes out all the data with bit masks and other auxiliary 
data. That is, rather than having just m/z vs. intensity data arrays, 
there are extra data arrays carrying more info. Example:
        m/z     Intensity       Monoisotopic?   S/N     Charge ...
        501.00  1000            Yes             50      3
        501.33  500             No              20      3
        501.66  100             No              10      3
        751.00  5000            Yes             150     2
        751.50  2000            No              70      2

Option (1) has the advantage of being very simple for any software 
consuming mzML to interpret the data arrays without any ambiguity. The 
disadvantage of option (1) is that any change in data requirements 
requires that one go back to the original raw data and re-generate a 
completely new mzML file. By contrast, option (2) has a lot more data 
embedded in the file and is much less likely to require re-generation of 
mzML, but requires more intelligence from the software consuming the mzML 
- for example, if only the monoisotopic peaks are desirable, the consuming 
software must nevertheless understand that it can't just read the m/z and 
intensity data arrays but it MUST ALSO read the monoisotopic data array 
and throw out some data values based on the contents of the monoisotopic 
bitmask array; otherwise the results can be bad.

Is there any guidance for how these things should be handled in mzML? What 
are the assumptions made by existing mzML consumers?

Thanks,
Wilfred