There are three kinds of community annotation that have been talked about:

Bulk annotation upload

Lots of people have GFF/BED/etc. files that they'd like to see rendered. We ought to make this dead simple. Existing examples include Santa Cruz's upload functionality:
http://genome.ucsc.edu/cgi-bin/hgCustom
and GBrowse's:
http://gmod.cvs.sourceforge.net/*checkout*/gmod/Generic-Genome-Browser/htdocs/annotation_help.html

As I see it, the fundamental difference between "annotation upload" as currently implemented and "community annotation" like we're talking about is that with annotation upload the uploader is the only one looking at their annotations, while with community annotation potentially everyone else is looking at them as well.

This immediately raises a number of questions, like how to manage all the uploaded tracks. These will obviously vary in usefulness, and we don't want to clutter up the display too much; on the other hand we don't want to make people pick from a huge list as a prerequisite to using the browser. This suggests that we'll want to have a list of "blessed" or "default" tracks, but then you have the problem of choosing those. Or we could have a track ranking mechanism where we show the top n highest ranked tracks, but then we have the problem of coming up with the track rankings. One option would be a moderation system (e.g., "Digg this track [+/-]"). In the fancy extreme we could even have a collaborative filtering mechanism where we "recommend" tracks to people based on their previous moderations.

In any case, making annotation upload work is a prerequisite to everything else, so that's the first priority IMO. The main things we have to do differently from current annotation upload are kicking off a pre-rendering or primitive database fill operation and coming up with some way of notifying the user when their track is ready to view. Anything else?

Individual feature creation/editing

It would be neat to be able to "draw" and "resize" features by clicking and dragging with the mouse. I believe this is possible to do in the browser, but it is fairly fancy and might be difficult to do in a cross-browser way. Suzi Lewis and Chris Mungall at Berkeley BOP seem to think that this need is adequately served by fatter clients like Apollo. Can anyone think of a use case where someone would want to be able to adjust feature boundaries but would be a casual enough user to not want to install Apollo? I would imagine that most people who want to e.g. edit gene models would be involved enough to install some software but my experience here is limited. If no one can come up with a compelling use case for having this on a web browser I'd be tempted to leave this out entirely, or at least leave it until the other community annotation things are implemented.

Feature "commenting"

Not sure what to call this (annotation annotation?), but I'm sure there are lots of biologists out there who know a lot about a relatively small number of genes. This is the kind of knowledge that wouldn't be a good fit for creating an entire track with (so it doesn't fit well into the bulk annotation upload mechanism), but would work better for adding information about features on an existing track. Examples that I can imagine include adding GO terms to existing annotations, adding information about where a certain gene is expressed, what pathways a gene is involved in, literature references, etc. Some of these things could be stored in Chado (how much?) but in general the kinds of information that would go here are hard to predict IMO, so I'd like to have the option of storing it in something less structured.

On the unstructured side, we could have a wiki with a page for each gene, transcript, protein, whatever. One moderately-structured option would be to use a semantic wiki like Semantic MediaWiki. I like the idea of having this information available in RDF, and the URI-based named attribute and typed link structure fits fairly well with the DAS/2 feature/prop mechanism AFAICS.

One nice thing about a semantic wiki is that we could treat certain labels specially and update the relevant chado fields. For example, if someone puts a

[[go_term::ion binding]]

In the short term I'm thinking of generating a semantic wiki page for each annotation (directly from Chado or using Bio::DB) and having that be (one of) the link(s) in the feature mouseover popup. Per the discussion here from last November, I'd like to implement those popups with DAS/2, which ought to be reasonably straightforward to access from javascript.

-- MitchSkinner - 06 Feb 2007