[Biococoa-dev] more ramblings

[Koen van der Drift]

On Dec 3, 2004, at 12:03 PM, Alexander Griekspoor wrote:


>> I like the BCAnnotatedSequence name. Although it kinda leaves out the 
>> features.
> True, but in some way also in that case I like to think as the 
> sequence being annotated...
>>

Yes, that is true.


>> BTW, I'm still confused about what is the difference between a 
>> feature and an annotation. I think we should define very well what we 
>> want to store in either one. And if the BCFeature and BCAnnotation 
>> aren't objects, what do you intend them to be?
> Oh sorry, in the previous email you wrote:
>
>>
>> BCSequenceWrapper
>>       |
>>       -> has-a BCSequence, BCFeatures, etc
>>
>>
>> BCFeatures
>>       |
>>       -> has-a dictionary of features (d'oh)
>
> ..so I thought you wanted to create a BCFeatureS class, which would 
> have a dictionary of features (BCFeature?).

I don;t know - I thought that was similar to what John proposed last 
week. I really have no idea what the structure should be. It was just 
an example to get the discussion going.


> But I guess you meant the same thing as I do: A BCSequenceWrapper 
> (object) has a dictionary containing multiple BCFeature objects.
>

Yes.  Although we probably will use BCAnnotatedSequence which is a 
BCSequence, instead of has-a (see my other email).


> As for the distinction between features and annotations: it's pretty 
> simple, features are linked to certain symbol ranges (phoshorylation 
> at 12, helix from 15-18), annotations are like metadata, author = 
> koen, link to article = ... , creation date.. etc

Yes, that makes sense.

>
> I think that both should be tagged by string identifiers (like 
> BCFeatureTypeAlphaHelix and BCAnnotationTypeAuthor), and we can 
> predefine a complete set of them to be sure that things like author, 
> date etc are standard items. The main advantage is that a developer 
> can add it's own tags to for instance save a marked region in a 
> sequence: (AGFeatureTypeMyMarkedRegion). We will make sure that these 
> get written in and out of files for him, so he doesn't have to invent 
> his own filetype.
> Note that it should be no problem to add multiple instances of 
> BCFeatures with same tag (like authors).

Sounds like a good plan. Not that this will not work with a dictionary, 
because it can have only one of each key. BTW, I'm pretty sure that 
there is a xml scheme out there that tries to unify these things for 
various dataformats. We could consider adapting such a thing (if it 
exists).

>
>>
>> Should we also make a special class BCAnnotatedSequenceReader that 
>> takes care of extracting as much as info as possible from a file? Or 
>> should we do everything with BCSequenceReader.?
>
> Good question, I don't know, there should definitely be the option to 
> either read in the features and annotations or not. It can go into one 
> class, but it's perhaps nicer to have to classes. Then again, it would 
> require a lot of code duplication which we don't want. We could make 
> the one a subclass of the other, the annotated one takes off where the 
> super class stops. No idea...

Food for thought.


cheers,

- Koen.