[Bioperl-l] Phyloxml SeqI without actual sequence

[Chris Fields]

On Jul 29, 2008, at 12:59 PM, Hilmar Lapp wrote:

> On Jul 29, 2008, at 1:31 PM, miraceti wrote:
>
>> sometimes the clade looks like.
>>
>>              <sequence>
>>                 <annotation>
>>                    <desc>alcohol dehydrogenase</desc>
>>                    <confidence type="probability">0.99</confidence>
>>                 </annotation>
>>              </sequence>
>>
>> sometimes the clade looks like.
>>
>>                <sequence>
>>                 <symbol>ADHX</symbol>
>>                 <accession source="UniProtKB">P81431</accession>
>>                 <name>Alcohol dehydrogenase class-3</name>
>>
>> <mol_seq>TDATGKPIKCMAAIAWEAKKPLSIEEVEVAPPKSGEVRIKILHSGVCHTD</mol_seq>
>>                 <annotation ref="EC:1.1.1.1"/>
>>                 <annotation ref="GO:0004022"/>
>>              </sequence>
>>
>> Do you think it's better not to create the SeqI in the first case?.
>
> It's an option but to me it threatens to lose the fact that the  
> annotation properties pertain to a sequence, not to something else  
> (such as the tree node, for example). So, if you don't attach the  
> annotation to a sequence object you create, what would you attach it  
> to then? In the absence of a good answer creating the sequence  
> object may the only way.

This is something we had discussed on IRC fairly recently (annotation  
that could be considered specific to the Tree, Node, Sequence, or  
combinations thereof).  From the documentation the only thing  
designated specifically with 'annotation' is the sequence, but there  
are several complex types which don't easily fit within a set  
AnnotationI implementation (such as <clade_relation>/ 
<sequence_relation>) or within class data.

Mira, are there some examples where Trees/Nodes/Seqs could reasonably  
be expected to have separate annotation?

> I thought I'd recall some kind of Bio::Seq::EmptySeq object for  
> exactly this kind of case, but apparently there isn't one there, so  
> maybe I was just dreaming this up. However, a sequence object that  
> is optimized for this case may be useful to have, though thinking  
> about it, I'm not exactly sure what overhead the sequence slot  
> really adds to all the other overhead added by feature and  
> annotation slots. The only thing that really comes to mind is that a  
> sequence object with a zero length sequence won't allow you to add  
> features.

I like Jason's suggestion of using a flag (maybe specifically calling  
for -emptyseq => 1 to prevent warnings).

> Sorry for this random stream of thoughts here, I hope I'm not  
> totally wasting everyone's time here.
>
> 	-hilmar

Not at all.  That's what this is all about!

chris