Re: [Psidev-ms-dev] mzML's binary arrays

[Matthew C. <mat...@va...>]

Hi Marc,

I admit I was hasty to say there was no difference. As you point out, 
the CV way makes it a "categorized" comment wheres the userParam is 
totally uncategorized. I still think a special term for such a comment 
is counter-productive for encouraging inter-compatible software.

Instead, a small modification to your software would allow you to 
enumerate all the userParams and cvParams in the array and output them 
in name-value pairs. So you'd have:
array 0:
  name="array name" type="array type" units="signal-to-noise ratio"
  min: 234
  max: 435435
  avg: 4545

Putting in "categorized" comment terms is a slippery slope IMO. It would 
lead to other similar terms in other places and ultimately we'd be like 
mzData with little or no control over the values of string variables, 
doom and gloom notwithstanding. ;)  At least is could lead to 
implementations of mzML that are widely incompatible.

As for 528 parameters across all your tools, how many of those are just 
for your data processing algorithms (that take raw data mzML as input 
and could write processed mzML as output, as opposed to a search engine 
which would write pepXML or analysisXML)? For dataProcessing, I at least 
would prefer algorithm-centric terms instead of tool-centric terms since 
many tools could implement the same algorithm and more importantly, many 
tools will run multiple algorithms. So we would have a term for boxcar 
smoothing, Savitzky-Golay smoothing, etc. Some algorithms might be 
coupled tightly with proprietary tools (e.g. Mascot Distiller or Protein 
Pilot's peak picking) but we can still call it something like the 
"Protein Pilot Peak Picker". :)

-Matt


Marc Sturm wrote:
> Hi Matt,
>   
>> I can't think of any notable semantic difference between userParams and 
>> cvParams with uncontrolled string values. In both cases, you would have 
>> to deal with an extra term that only your algorithm and its downstream 
>> users know about. In both cases, the extra algorithm-specific arrays are 
>> unusable for other tools (except ones you make of course or that are 
>> made specifically to work with it). In both cases, the uncontrolled 
>> string value cannot be relied on except in very controlled circumstances.
>>   
>>     
> That's not entirely true in my opinion. We have a small command line 
> tool that can display statistics about these arrays.
> Without a defined way to give the array a name the statistics would look 
> like that:
>
> array 0:
>   min: 234
>   max: 435435
>   avg: 4545
> array 1:
>   min: 234
>   max: 435435
>   avg: 4545
>
> With a defined way of naming arrays it would look like that:
>
> array 'some descritpion of the array content 1':
>   min: 234
>   max: 435435
>   avg: 4545
> array 'some descritpion of the array content 2':
>   min: 234
>   max: 435435
>   avg: 4545
>
> At least to me the second alternative looks much better. Of cause we can 
> store the name in the userParam 'name'.
> But other tools would store it in the userParam 'Name' or 'custom_name' 
> or 'custom name' ...
> I really think there should be a controlled way to give an array a 
> user-defined name.
> There was a way in mzData (optional XML attribute 'name') and it's a 
> step back not to have one in mzML.
>
>   
>> However, if your peak picking algorithm is versioned, it's exactly the 
>> kind of thing we want in the CV. We want a term to briefly describe the 
>> algorithm (which would go in dataProcessing) and also terms to describe 
>> the parameters that a user can set. At the same time, CV terms for your 
>> custom extra arrays could be added as well.
>>   
>>     
> We'll compile a list of the TOPP tools with short descriptions and post 
> it on this mailing list.
> The parameters will most likely not be included as the current count of 
> parameters for all tools is 538 and they might change from release to 
> release.
>
> Best,
>   Marc
>
> -------------------------------------------------------------------------
> Check out the new SourceForge.net Marketplace.
> It's the best place to buy or sell services for
> just about anything Open Source.
> http://sourceforge.net/services/buy/index.php
> _______________________________________________
> Psidev-ms-dev mailing list
> Psi...@li...
> https://lists.sourceforge.net/lists/listinfo/psidev-ms-dev
>
>