Information visualization confronts abstract data by graphically representing relationships between pieces of data, and visually bringing out features inherent in the data. We often encounter multivariate data, but with only a few dimensions to explore on screens with limited resolution, we must think of creative ways to depict the data. Previously, we presented a new technique for representing information contained in biological sequence similarity search reports that maps three of the most important variables--- position, score, and frame---to the screen [6].
Seeing the potential of information visualization, the biologists on our research team asked for ways to visualize other information in the data. Responding to this need, we developed a technique that allows the user to arbitrarily map any of twelve variables onto the X, Y, Z, and time axes. We showed how biologically significant features can be investigated with the additional time axis. Moreover, we showed how the user can develop simple queries on this data using visual query filters, and how the filtering reduced the clutter in the information space. The case studies showed the power of the combined technique in finding, extracting, exploring, and analyzing features that were hard to find previously. We showed how the added time axis and visual query filters provided an effective way to analyze a plant sequence. We also showed the history of the HIV sequence by visualizing the increase in HIV related sequence information over the years using AlignmentViewer. The combined approach of arbitrary mapping of variables to axes, animation, and visual query filtering makes exploration in the information space more productive.
The HIV history animation and other related information on AlignmentViewer can be found at AV's home page (http://www.cs.umn.edu/~echi/av.html). In addition, AlignmentViewer is in daily use by the biologists, and 45,000 AV visualizations can be found in the similarity reports of plant genome sequences at our project's home page (http://lenti.med.umn.edu/general_cdna/wais_search.html).
There are many possible directions for further work. One possibility is to link the database to the visualizer directly to explore different interfaces between the visualizer and the database. A second possibility is to use more powerful and flexible filtering techniques. An additional possibility is to explore ways to visualize multiple search reports simultaneously.
We have extended AlignmentViewer with a set of powerful visualization techniques, based on feedback from the biologists. By decoupling the variables and allowing these variables to map to any of the three spatial and the one temporal axis, AlignmentViewer enables molecular biologists to study the intricate relationships between similar sequences.