[DAS] Trellis/Ivy: DAS1 --> DAS2 transformational proxy

[Gregg Helt]

If anyone wants to test the DAS1 --> DAS2 transformational proxy with a
DAS/2 client, you can try it with the Integrated Genome Browser
(IGB)<http://www.affymetrix.com/partners_programs/programs/developer/tools/download_igb.affx>by
adding this line to your igb_prefs.xml file (usually located in your
home
directory):

<das2_server name="Trellis/Ivy DAS1 proxy" url="
http://www.genomancer.org/das2/das1_proxy/sources?type=Chromosome&capability=types"
/>

Note that this server will show up in the "Data Access" tab for the current
genome and the "Pick Genome" button for genome selection along with other
DAS/2 servers, instead of in the "Access DAS1 Servers" menu item.

There are a few caveats.

IGB assumes that a DAS2 server provides "features", "segments", and "types"
capabilities for all the sources/versions it serves up.  But there are lots
of DAS servers in the registry that don't have a straightforward way for the
proxy to transform and provide all three capabilities.  To try and minimize
the number of sources IGB will have trouble with, the sources query above
includes filters to restrict the response to a (transformed) subset of the
sources listed in the registry that are likely to meet this requirement.
But there's currently no way to filter for exactly what is wanted (multiple
"capability" filters are OR'd together, not ANDed), so be warned that there
may still be some sources in the response that IGB can't deal with.

The other major caveat is that in order to compare annotations transformed
via the DAS1 proxy server alongside annotations from other DAS/2 servers,
IGB needs to be able to tell that they are using the same coordinate system
(genome assembly).  Currently the DAS1 registry and DAS/2 use different
identifiers for the same genome assembly.  Hopefully they can start using
the same identifiers soon, so I didn't really want to add a short-term (and
brittle) complete id mapping strategy in the proxy.  Until DAS1 and DAS/2
use the same coordinate identifiers the mapping at least for IGB can be done
in IGB, but to do this for the main Affymetrix IGB web start deployment
requires modification on the NetAffx Affymetrix web site (Steve, forgot to
talk to you about this yesterday -- I'll send you details).  So to test for
the next day or two I've done a single hardwired coordinates ID mapping in
the proxy for the latest human genome assembly (NCBIv36).  If you stick to
this genome in IGB you should be able to overlay data from the Trellis/Ivy
DAS1 proxy and other DAS2 servers.

Also keep in mind that none of the performance optimizations that the
Affymetrix Genometry DAS/2 server uses are implemented yet in the
Trellis/Ivy proxy.  So it's best to try out the proxy on small regions of
the genome.

Please let me know if you try this out how it works, any bugs you find, etc.

       Gregg

On Tue, Oct 28, 2008 at 11:02 AM, Gregg Helt <gregghelt at gmail.com> wrote:

> I'm pleased to announce I've got a DAS1 to DAS2 transformational proxy
> working.  I'm calling it Trellis/Ivy.  Trellis is the overall DAS2 server
> framework, and Ivy is the DAS1 proxy plugin.  The source code is available
> via SVN at http://code.google.com/p/genomancer/ .  This should definitely
> be considered a beta release, there are plenty of features that aren't yet
> implemented and I'm sure still some bugs in what is implemented.
>
> I'm hoping a stable version of this can eventually be set up close to or
> integrated with the DAS1 registry.  For current testing I've deployed an
> instance of Trellis/Ivy at genomancer.org in the Amazon EC2 cloud.  It's
> hardwired to proxy for the Sanger DAS1 registry<http://www.dasregistry.org/das1/sources>.
> Some example DAS2 queries and the DAS1 queries they proxy for:
>
> All DAS1 registry sources:
> http://www.dasregistry.org/das1/sources
> All DAS1 registry sources --> DAS1+2 sources
> http://www.genomancer.org/das2/das1_proxy/sources
>
> DAS1 filtered sources
> http://www.dasregistry.org/das1/sources?type=Chromosome&capability=types
> DAS1 filtered sources --> DAS1+2 filtered sources
>
> http://www.genomancer.org/das2/das1_proxy/sources?type=Chromosome&capability=types
>
> Ensembl DAS1 entry points
> http://www.ensembl.org/das/Homo_sapiens.NCBI36.reference/entry_points
> Ensembl DAS1 entry points --> DAS2 segments
>
> http://www.genomancer.org/das2/das1_proxy/www.ensembl.org/das/Homo_sapiens.NCBI36.reference/entry_points
>
> Uniprot DAS1 types
> http://www.ebi.ac.uk/das-srv/uniprot/das/uniprot/types
> Uniprot DAS1 types --> DAS2 types:
>
> http://www.genomancer.org/das2/das1_proxy/www.ebi.ac.uk/das-srv/uniprot/das/uniprot/types
>
> Uniprot DAS1 features (from UniProt DAS home page example)
> http://www.ebi.ac.uk/das-srv/uniprot/das/uniprot/features?segment=O35502
> UniProt DAS1 features --> DAS2 features
>
> http://www.genomancer.org/das2/das1_proxy/www.ebi.ac.uk/das-srv/uniprot/das/uniprot/features?segment=O35502
>
> Ensembl DAS1 features (from Ensembl DAS home page example)
>
> http://www.ensembl.org/das/Homo_sapiens.NCBI36.transcript/features?segment=13:31787617,31871806
> Ensembl DAS1 features --> DAS2 features
>
> http://www.genomancer.org/das2/das1_proxy/www.ensembl.org/das/Homo_sapiens.NCBI36.transcript/features?segment=13;overlaps=31787616:31871806
>
> UCSC-Gencode segment/overlap/type-filtered, URL-encoded, DAS2 feature
> request (copied from IGB testing):
>
> http://www.genomancer.org/das2/das1_proxy/hgwdev-gencode.cse.ucsc.edu/cgi-bin/das/hg18/features?segment=http%3A%2F%2Fwww.genomancer.org%2Fdas2%2Fdas1_proxy%2Fhgwdev-gencode.cse.ucsc.edu%2Fcgi-bin%2Fdas%2Fhg18%2F21;overlaps=26014828%3A26071595;type=http%3A%2F%2Fwww.genomancer.org%2Fdas2%2Fdas1_proxy%2Fhgwdev-gencode.cse.ucsc.edu%2Fcgi-bin%2Fdas%2Fhg18%2FknownGene
>
>
> Basic DAS1.53 query --> DAS2.0 conversions not yet implemented:
>     DAS1 sequence --> DAS2 segments
>     DAS1 dsn --> DAS2 sources
>     DAS1 stylesheet --> DAS2 stylesheet
> I'll post a summary of plans for these queries and other features soon, but
> I wanted to at least send out this announcement so people who are interested
> can check out the examples.  Please let me know what you think.
>
>     Gregg
>
>
>
>