Web-apollo-meeting-2011-3-15

3/15/2011: Suzi, Gregg, Ed, and Nomi

Progress Report

• WebApollo year 1 progress report submitted. Thanks to all who contributed to it.

Hardware

• What hardware do we need for a WebApollo server?
  • One server (Linux or OSX). (Budget?)
  • Server itself doesn't need to be too studly--mostly we need a lot of storage.
  • Four 3-terabyte drives (or at least 8 terabytes to start with)
  • Need space to double every year (but hopefully that will be constant cost/year as price of memory drops)
  • Server speed is not a huge issue. The client side (how much data a browser can handle) will be the limiting factor.
  • Seth says we have space on campus and the networking we need.
• Action items:
  • AI: Suzi will check on time constraint (when does this grant year end? April 30?) and hardware budget
  • AI: Seth will get a price quote
  • AI: Seth and/or Ed will do the ordering
  • AI: Ed will get Tomcat up on the server
  • AI: Gregg will deal with the rsync
• Backups and server use by outside groups
  • Need to back up user data (though not the UCSC data, which we can always get again).
  • Larger groups may set up their own servers, but we need to be a server for smaller groups that want to use WebApollo.
  • Would we have to set limits about how much data users could save on our server?
• Seth proposed buying a server with the following (which we can afford with our current hardware budget):
  • 2x 4 core
  • 6GB RAM
  • 7x 2TB hd (RAID 6 + one hot spare = 8TB)
  • 1x 120MB SSD boot disk
  • 3U profile, 12 hd slots
  • 4x Gb ethernet (for bonding on two interfaces)
  • Gregg said: "Seth and I talked briefly in person on Tuesday, and he convinced me that it makes sense to spend the money for enough hard drives to set up RAID 6 for the UCSC data. And I agree to allow for future expansion we should go ahead with a 4U 16 slot box."
  • I believe the server has now been ordered?

JBrowse issues

• JBrowse currently only supports two-level feature hierarchy.
  • Mitch wants to do a JSON approach that allows arbitrary levels of features, and also different kinds of features at the same level (possibly with different fields).
    • Gregg and Mitch have talked about details of this approach, sounds like it'll provide what we need
    • Mitch may be able to implement this in the next several months
    • If Mitch doesn't get to it before we need it, Gregg can implement

What does WebApollo need to provide?

• We looked at old Apollo to see what it has so we could decide what WebApollo needs.
• Authentication service with read/write privileges (records who made the annotation)
• Sequence window (old Apollo)--view as cDNA, AA, etc., and can copy&paste or save to a file
  • Suzi notes that it's really annoying when web apps pop up extra windows
  • Gregg says what he'd want is a menu item, Copy to Clipboard
• Alignment viewer
• (Exon Detail Editor--don't want)
• Show start/stop codons (when you've zoomed in enough to see bases)--calculate on the fly on client. (Would need to show which strand they were on.)
• Stuff that's in Apollo's Annotation Information Editor--what do we need?
  • Ability to rename annotations
  • Add/edit comments
  • Synonyms
  • Dbxrefs (JBrowse already provides this?)
  • Ability to paste in new exon boundaries (?) (Maybe this doesn't belong in the Annotation Info Editor)
  • Little pieces of info, some of which are calculated automatically (e.g., missing start/stop codons)
  • Ontology terms? (Old Apollo doesn't really have this, except via the dbxref, which has a checkbox for "is from an ontology".)
    • Gregg suggests that this panel could pull out as a slider, or it could be brought up by a right-menu option. We don't want it around taking up space unless it's requested. We don't want it to pop up as a separate window.
• Ability to create de novo annotation (Apollo offers a right-menu option to choose annotation type)--types should be a subset of the Sequence Ontology, but you should also be able to create your own new annotation type.
• Time-stamp annotations
• Let user set an anchor? (Keep a point centered even when zooming in and out.)
• Should we offer zoomed-in and zoomed-out views (vertically) in the same screen? Suzi says yes. Can have two vertical scrollbars. Can zoom the two views independently.
• Splice view (like in IGB)--show all the exons (plus a bit of the introns) and leave out the boring empty intron stretches.
• Annotation editing operations:
  • Undo
  • Merge
  • Split
  • ...

Discussed commonality between editing operations

• Interval algebra
  • Add exon, extend exon, merge exons, create transcript, merge transcripts can all be considered union operations on exon/transcript intervals
  • Delete exon, shorten exon, split exon, delete transcript, split transcript can all be considered intersect (or alternatively XOR or NOT) operations on exon/transcript intervals
  • Not exactly sure what this implies for user interface, but hopefully thinking about this will help simplify the GUI

Undo/Redo

• Discussed global undo vs. per-annotation undo
• Implementing undo/redo as ancestor/descendant traversal
• Handling splits and merges -- implies multiple ancestors, multiple descendants
• UI for undo/redo -- back/forward, select from version list, or both