[DAS2] current status of DAS

[Andreas Kahari]

On Sun, Sep 24, 2006 at 08:17:38AM +0000, Rowan Kuiper wrote:
> I have a few questions. I’m a bioinformatics student and for an internship 
> at the Erasmus University in Holland I have to investigate the current 
> status of DAS. I’ve been trying to work with DAS a couple of weeks now and 
> the impression I get is that it is a bit messy. Perhaps this is because I 
> don’t understand DAS very well and can you explain it to me.

DAS is a specification of a communication protocol originally intended
to provide a web service for serving GFF-like annotation data.

> - First of all, will DAS2 ever be finished. I saw on  the biodas site that 
> the 2 year  development started in 2004. But when I looked at sites that 
> should propagate the development, DAS seems to be out of focus. You think 
> DAS is still alive or is there something else that took its place?

The stagnation of the DAS/2 developments that you refer to is outside of
what I know very much about (the frequent telephone conference mailings
on this list suggests it's not stagnated at all).  I work full time with
a large number of research groups who are using DAS/1 as a tool for data
integration in various ways.  So in Europe, at least, DAS/1 is very much
alive.  Also, within the Ensembl Genome Browser (www.ensembl.org), more
things are done through DAS than what you might think.

> - Why don’t all servers support all commands. Some reference servers for 
> example don’t support the entry_point command. How do I request features 
> when I don’t know which segments the server contains? I imagine that these 
> great differences in how to use different servers could be very problematic 
> when implementing a viewer.

Lazy maintainers, possibly?  Could you please provide us with concrete
examples of these reference servers?  If any of them are within my
control, this would give me a chance to fix them.

> - It seems that the only way to retrieve information from a server is to do 
> a request for a certain region. Is there a way to ask for a specific 
> features.

This is an artefact of the way genomic annotation viewers work.  They
provide the user with a view of a genomic region at a time.

According to the specification (DAS/1), the 'features' request may be
tailored to only return certain feature IDs on a given segment using the
'feature_id=ID' argument.

Whether this capability is implemented by a particular server or not
should be evident from the HTTP headers sent back from the server.

Again, since people are lazy (me too), and since clients never, as far
as I am aware of, make use of this capability, it is seldom implemented.

> - Is the Sanger Registry Server reliable or is it something of the past? It 
> would be very nice if all available sources where listed there but just a 
> small part of the sources I found where in the list.

I'll leave this one for Andreas Prlic.

> - When I have to serve data that needs some extension on the XML structure, 
> would it be a problem to just do it. How would clients handle these 
> extensions. Ignore them or somehow parse them?

You're free to add whatever XML you feel a need to add.  A well behaved
DAS client will ignore it.  If the response still contains the necessary
bits and bobs, then it is in my opinion still DAS, otherwise you've
broken the protocol and the response will be unusable by any existing
client.  There is no magic in clients that will tell them to look for
XML structures that are not specified as being part of the DAS response.

> - And last, one of the goals of DAS is to be able to integrate biological 
> data. When I for example want to compare my data to EnsEMBL features I will 
> have to set up my own server that serves features referenced to the same 
> genome as the EnsEMBL features. So I wonder if there exists reference 
> servers that contain the current genomes  of EnsEMBL, NCBI or UCSC.  I found 
> http://das.ensembl.org/das/ensembl_Homo_sapiens_core_38_36 which always 
> replies an out of memory error even with the entry_points command and 
> http://das.ensembl.org/das/ensembl1834 which seems to work properly but its 
> transcript server ens1834trans also returns out of memory errors.

If you wish to do numerical (not visual) comparisons of data against
Ensembl, I believe this would be easier with the help of the Ensembl
Perl API.

Ensembl nowadays serve reference sources from the www.ensembl.org/das
server.  See Eugene's reply for examples.

> I think that there are some people here that can tell me their view on the 
> subject.
> Thanks in advance,
> Rowan Kuiper

Regards,
Andreas

-- 
Andreas Kähäri
Ensembl Software Developer
European Bioinformatics Institute (EMBL-EBI)