Dalliance

Thomas Down

http://www.biodalliance.org/

Genome browsers

• Can we have the best of both worlds?

Technologies align

• Solid Javascript implementations
• Rich browser-based graphics (SVG/Canvas)
• Browser vendors focussing on performance (games!)
• DAS: Distributed Annotation System
• CORS: Cross-origin Resource Sharing
• HTML5

Let's try it out

What is DAS

• The distributed annotation system
• Simple HTTP queries to retrieve sequence annotation

   GET http://server/das/xyz/features
          ?segment=22:30000000,40000000

   <FEATURE id="...">
      <START>30000000</START>
      <END>30000001</END>
      <TYPE id="density">read depth</TYPE>
    </FEATURE>

• Plus stylesheets, registry services, etc.

What is CORS?

• Javascript data fetching: XMLHttpRequest
• Same-origin policy
• Origin ~= Server
  • e.g.: http://www.biodalliance.org/
• CORS: server can declare a resource suitable for cross-origin requests
• Simple examples:

  Access-Control-Allow-Origin: *
  Access-Control-Allow-Origin: http://biodalliance.org

• Popular DAS middleware does this for you.

More data...

• Data volumes increasing
• More datasets
• Larger datasets
• Experiments like:
  • ChIP-seq.
  • Exome resequencing.
  • ...
• Small labs produce genome-wide data
• Few resources to help get data out

How does this fit DAS?

• DAS XML has served us well
• Relatively rich data model (IDs, links, notes)
• ...but XML is verbose

   <FEATURE id="why_oh_why?">
      <START>30000000</START>
      <END>30000001</END>
      <TYPE id="density">read depth</TYPE>
      ...and more...
    </FEATURE>

• Requires a server deployment
• ...and someone to deploy that server
• Relational DB backends don't always scale

The binary solution

• Dense binary formats
  • bigbed: general features
  • bigwig: dense quantitative data
  • BAM: short read alignments
• All have indices
• Therefore, random access is possible
• HTTP supports random access using Range
• .htaccess file:

  Header set Access-Control-Allow-Origin "*"
  Header set Access-Control-Allow-Headers "Range"

Binary Dalliance

• Recent browsers have bearably-good APIs for binary data
• We can use these to directly support bigwig and bigbed and BAM files
• ...demo...

Dalliance in the future

• More data!
• More formats? (please ask!)
• More interactivity
  • Jump around quickly between interesting regions.
  • Link to other tools.
  • Connect with other biologists.
• Easier setup: complete browser from static files
• Alignments (assembly mapping; comparative genomics)
• Performance (SVG -> Canvas?)

Acknowledgements

• Tim Hubbard
• DAS Developers
• Dalliance testers



• My other life: Wellcome Trust RCDF


http://www.biodalliance.org/