Skip to content

user_scenarios

Jump to bottom Edit New page
David Osumi-Sutherland edited this page Mar 23, 2014 · 14 revisions

The aim of this document is to outline realistic scenarios for users of VFB, to assess the extent to which the site lets them achieve their aims and to suggest improvements wherever VFB falls short.

Scenario 1:

A researcher has found a neuron with terminals in some poorly mapped part of the brain that she hypothesises might function in perception of auditory signals. She wants to know:

  1. What is this region of the brain called and what is known about it from the literature? Solution: Double click on the Brain region in the stack browser - note, using Wedge as a running example. This provides a name, and image and a description along with definitional references and a set of options for queries to find further information.

    • POTENTIAL PROBLEM: The researcher cannot easily relate their preparation to the parts of the brain stack or the brain regions as defined in VFB, since they have not been able to align it to a standard brain.
      • Suggested improvements:
        • Offer registration to preparations stained with common antibodies?
        • Provide a guide to registration?

  2. What is know about the function of this brain area in the literature?

    • Can we find all papers that mention e.g. parts of this brain region?
    • Solution (not ideal!): Query for parts and for overlapping neurons. Trace back papers from these.
      • Suggested improvements: All anatomy pages should have a complete reference list underneath, with some indication of what data the paper has: anatomical; expression; phenotype. This should probably be split into direct mentions (or perhaps direct + parts) vs indirect (including refs for overlapping neurons). The challenge of this is that references need to be pulled from both the ontology and multiple types of annotation query. This is not achievable with the current live architecture, but would be relatively easy with pre-computation into SOLR.

  3. NOTE: The way that the following question is phrased suggests that the neurone has already been ascribed to a known class. Achieving this should therefore come before asking whether anything is known about it? Or is this what you actually mean by the first question?

  4. Is anything known about this neuron:

  • From the literature?
    • Solution: query for neurons with some part in this region
    • Problems: There may be many neurons with parts in the region. The results list doesn't provide thumbnails (even when these are available). So a user has to check each neuron individually. If no images are available on the term info page for the neuron, she has to download the original papers to check.
    • Suggested improvements:
      • Display thumbnail images in results lists when querying for neurons/anatomical structures.
      • Make the collection and assignment of images to neurons a high priority for data collection.
  • From bulk data?
    • Solutions:
      • query for images of neurons with some part in this region, clustered by shape. This has the advantage that the report is quite short, due to clustering, and comes with nice clear thumbnail images. Clusters can be further explored as 3D rotatable webGL. Members of clusters can easily be viewed, and these come with classification (if present - coverage is currently patchy) and information about transgene expressed and other regions overlapped. Any individual neuron can be viewed on the stack browser, in combination with other neurons or transgene images if needed.
      • Query for images of neurons with some part in this region, unclustered. This has the disadvantage that the list is long, but the advantage that it is more comprehensive. All results come with thumbnails and typing, if known.
      • Problems: Very limited mapping of neuron images to types outside of outside of antennal lobe neurons.
        • Suggested improvements: Make mapping of neurons to types a priority. Some of these could be done quite trivially with a bit of effort. Many could potentially be crowdsourced from the community.

Cahir: YES. However: (1) Can this be done at all on a large scale without community involvement? I don't think that we yet have a manageable strategy to produce a definitive thumbnail or larger images for every single neuronal class. (2) While Flycircuit have the kind of images (or WebGL derivatives of them) that would be good for this, at present there is no easy way to relate these to transgene expression patterns. Furthermore, transgene expression patterns are what most people will now use as the working definition for a neuronal class 95-99% of the time (although not quite 100%), and is what most people will want to know; and so mapping of neurons to types should as far as possible be based around expression patterns (I'm aware of the problems in practice, as you describe elsewhere here)

Lineage - Make this a separate scenario? And after the transgene scenario?????

  • Can I match this neuron to a lineage?
  • Are any tracts are known to innervate this region good candidates for carrying my neuron of interest?
    • Solution: Query for tracts that innervate this region.
      • Problems: No images - no painted tract stack is currently available on VFB, although there is one in the BrainName raw image data.
        • Suggested improvements: Add painted tract stack from BrainName. (Disadvantage is that this is a half brain, not good for registration).

  1. Do any known transgenes drive expression in this neurone?

  2. Is this neuron labeled by any known transgenes? (Suggestion from Cahir, I think this is closer to the way that most users think?)

    • Solutions:
      • If the neuron is among published classes, a single click lists transgenes known to be expressed in this neuron.
      • Search for transgenes with some parts in this region. Results include images where available. These are linked to their source and can be viewed on the stack browser - in combination with single neuron images if need be.
      • Problems:
        • Only expression patterns from bulk datasets have images associated:
          • Suggested improvements: Curate more images - from users and lit.
        • Even when images are available, going through each candidate and trying to work out if they might cover a neuron of interest is slow and hard;
          • Suggested improvements: Automated matching between single neuron images and expression patterns.
        • To get really specific staining, we need to make intersections between driver lines.
          • Suggested improvements: More curation of lines suitable for intersectional work (e.g. LexA, & split GAL4; Automated matching between expression patterns (basic overlap should be easy, but may give many false +ves).

  3. Can I find potential circuit partners for this neuron?

Add a custom footer
Add a custom sidebar
Clone this wiki locally