[Psidev-ms-dev] mzML comments

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

I recently looked over the mzML format draft and have a few comments.


Regarding the schema (http://www.sbeams.org/tmp/mzML0.99.12.html):

* For the "cvParam Mapping Rules," I suggest that the number of "musts" be 
decreased significantly (or changed to "mays"). For example, for the 
element <sourceFile>, there's a rule saying: "MUST supply a *child* term 
of MS:1000561 (data file checksum type) one or more times." This sort of 
information does not seem to be critical for downstream processing of mzML 
files and thus doesn't deserve "must" status. There are quite a few 
similar examples.

* For the elements <fileDescription>, <sourceFileList>, <sourceFile>, 
etc., would it make sense to change the name to be more general (such as 
dataSource) to reflect the fact that not all instrument data is stored in 
files? For example, for the Applied Biosystems|MDS Sciex instruments, some 
instruments (such as the QSTAR or QTRAP systems) store data in .wiff 
files, while other instruments (such as the 4700 and 4800 systems) store 
data in an Oracle database.


Regarding the controlled vocabulary (
http://psidev.cvs.sourceforge.net/*checkout*/psidev/psi/psi-ms/mzML/controlledVocabulary/psi-ms.obo
):

* Under "spectrum"-->"spectrum representation", the only 2 choices are 
"centroid mass spectrum" and "profile mass spectrum". That doesn't 
adequately capture the full range of spectrum representations. For example 
in addition to centroiding, the data could be de-isotoped, smoothed, 
converted to +1 charge, etc. Also, having just the 2 choices of centroid 
vs. profile is inconsistent with the software processing options listed 
under "data transformation"-->"data processing action", where a much wider 
range of options are listed ("baseline reduction", "charge deconvolution", 
"deisotoping", etc.). It seems desirable to expand the choices under 
"spectrum"-->"spectrum representation" to at least be consistent. 
Alternatively, maybe a slightly different categorization might make sense 
- something like raw data vs. processed data (full data vs. reduced data).

* Under "spectrum"-->"spectrum type", some triple quad-type scans are 
missing - precursor ion (scan Q1, fixed Q3), neutral loss (scan Q1 and Q3 
together with a constant difference between Q1 and Q3).


Please accept my apologies in advance if any of these topics have already 
been discussed/resolved previously, as I am new to this discussion.

Thanks,
Wilfred