[DAS] [Fwd: Re: Writeback implementation]

[Andy Jenkinson]

Gustavo Salazar wrote:
> Hi Andy,
> 
> The way I start to implemented is following the idea of interfaces that 
> MyDas has. In such a way i just extend the skeleton of MyDas to support 
> other kind of source called DasWritebackInterface, it allows that every 
> group implement the writeback in different ways but all of them should 
> provide the same capabilities(so far I define 3 methods, add, update and 
> delete)
> As the source class can implement more than one interface, my idea for 
> the query is that the source should implement both the writeback and the 
> annotation source and in that way the server will use the feature 
> command to provide the annotations.
> do you think that this way will be enough for the hypotetical case that 
> you describe?

I think this would be best, we just need to figure out how to clearly 
identify features from another source that are being annotated (see below).

> 
> I've been thinking that the commands in the writeback document should be 
> extended, because so far the defined commands are DELETES and UPDATES 
> (no command meaning ADD). An idea that i have is to put a command 
> SYNONYMS to tag 2 or more features that coming from different servers 
> and give the same information.
> On the other hand a control version for the annotations will be very 
> useful and in such a case at least 2 new command will be required 
> getHistorical to get all the meta-annotations done over one feature, and 
> setCurrent to define which version of the meta-annotations should be used.

I would suggest that ADD functionality can simply be provided by issuing 
an UPDATE but with a new feature ID. How did Asia do this? Likewise 
setCurrent (e.g. to rollback a change) could be done by the client, by 
reinstating the annotation with an UPDATE. I'd like to keep things as 
simple as possible, certainly for the client, as in my opinion this has 
been key for DAS in the past.

With regards to synonyms, this is a definite issue for DAS which is 
probably worthy of its own topic. This was proposed to be fixed via the 
METHOD tag (used to identify the "reference" server for the annotation), 
but I'm not sure if this has been adopted so far.

> By the way, Im not familiar with the Dasmi project, what is the idea of 
> those rankings?

The idea is that some servers provide interactions (which are 
annotations of proteins or domains), and others provide confidence 
scores for those interactions which are laid over the top. The 
documentation on dasmi.de explains how this works, but IIRC it relies on 
the IDs being unique across servers. This isn't guaranteed according to 
the DAS spec, so we will likely need to do something about that.