Many high dimensional exploratory statistical packages have the capability to mask or mark a subset of the data (see, e.g. [9]). These techniques reduce clutter by filtering, which makes navigation through the information space easier. The animation capabilities of our system provide one type of filtering, but we also provide additional slider-based filtering. Even though our system allows only four variables to be represented on the screen, we allow the user to specify a filtering range for any variable. This is done by the use of simple sliders for a high and low cutoff. In Figure 4, the user has specified to view only alignments that start from position 0 to position 399, with the length greater than 104 but less than 1787. The user can change the range on any of the twelve variables dynamically, and see the result immediately. Biologists can explore the information space interactively in real-time, seeing the result of the constructed query as they change the value of the high and low cutoffs.
Figure 4: Visual query filters for a simple query
This capability allows biologists to analyze different groups of alignments from an alignment report. Suppose a biologist wants to analyze an alignment report of a sequence 1000 residues long that contains many alignments along different regions of the sequence, and wants to view two regions of alignments separately. To analyze the first group, she simply filters out all alignments except the first region, say positions from 0 to 500. Then she can further examine these set of alignments by constructing animations, 3D scatter plots, or even further filtering. After examining this group, she can reset the filters to encompass only alignments between positions from 500 to 1000 for further analysis.
Our exploration with AV led to the implementation of a new system that allows the arbitrary mapping of variables to the four axes, where the mapping is specified by the user interactively. The user can animate the visualization or filter the result by building visual queries. In the next section, we present some interesting uses of this new system, and demonstrate how the decoupling of variables from the axes enables new types of analysis.