Interactive Pathway Interaction Data?
Hilary Spencer
Tuesday, 18 March 2008 20:48 UTC
Inspired by a Nodalpoint post and a related post on Nascent, I uploaded one of the NCI-Nature Pathway Interaction Database maps (ADP-ribosylation factor 1 pathway) to MapLib.
You can see the results here.
MapLib adds a Google Maps derived UI to arbitrary images, allowing one to zoom/pan the image and add location markers/popups. It also has a number of bugs, namely in adding HTML to one’s markers (I would randomly get “server errors”). There is a HTML-WYSIWYG textbox that doesn’t live up to its name (showing you a preview that doesn’t match the marker markup), and the marker placement is horribly inaccurate, shifting around depending on one’s zoom level.
Aside from the limitations of MapLib, this approach seems to some benefits over using a static GIF or JPEG (and even SVG). Yet, in some ways adding AJAX interactivity over a crummy graph is like adding new paint to an old rental apartment. Google Maps wouldn’t have been as successful if they had used poorly drawn maps (or even good maps that were just poorly adapted for online use).
Compare:
Paper map of Weimarer
Google map of Weimarer
The Google map was designed for zooming and panning from the beginning, where the paper map was not. Zooming out in Google Maps doesn’t just make the image smaller, but hides information not relevant at that level. Zooming to the city level filters out information about individual streets. Zooming to the country level hides information about all but the major cities. The Nodalpoint post cites the following article:
Zhenjun Hu, Joe Mellor, Jie Wu, Minoru Kanehisa, Joshua M Stuart & Charles DeLisi. Towards zoomable multidimensional maps of the cell. Nature Biotechnology 25, 547 – 554 (2007).
The authors’ key insight (IMHO) was noting that in order for a zoomable map of the cell to be useful, it would need to encorporate " semantic zooming, or changes in information density concordant with scale". My MapLib experiment fails in this area, which is probably why it won’t lead to a Nature Biotechnology publication ;)
Updated 18 March 2008 20:55 UTC
-
Replies
-
Well, maybe there’s not a Nature Biotechnology article in this experiment just yet, but its a good start. I’ve yet to find (or, more honestly, be told about) a good, adaptable map display system for biological networks & pathways. Maybe one is out there; my innards tell me that one should surely be possible.
Here’s a rough list of the features I’d like to see (and yes, I really would like to use these):
MUST-HAVES …
- Accept some initial form of graph layout e.g a network map generated by GraphViz.
- Zoom and pan — without adding/subtracting info depending on the display scale.
- A small navigator window showing the entire map, to orient the user when zoomed in.
- Mouse-over nodes and edges to show a pop-up info box (which should have text/HTML formating). Click on an object to do something outside the viewer (e.g. go to a web page or database record).
- Support different classes of nodes and edges (e.g. drugs, proteins).
- Search within the map (ideally flagging up all found terms visually in some way)
WOULD-BE-NICE:
- A magnifying glass – for mousing over complex map regions to see a local region zoomed-in
- Show or hide the various classes of nodes and edges
- Display only a portion of the map (not necessarily a rectanglar area – for which zooming would do), and to then sequentially add or remove nodes and edges from it.
- Pre-set views: a certain scale, region of the map, certain node/edge types shown, certain terms highlighted.
- Personalization. Save your own view settings. Tag items or views in some external annotation/tagging/bookmarking system.
- Change the visual display characteristics: e.g. the line width/color of different types of nodes/edges.
- Overlay separate maps, or comparing them side by side.
- Export maps for publication/further graphics processing.
That’s enough features for now, I think.
-
