Space Explorer is a Macintosh-based program for the visualization of high dimensional
spaces. Its means of representing such spaces is described in more detail
document is a very preliminary introduction to the user interface and some of
the early capability of the program.
Space Explorer is still in the early developmental stages. Although we have an
early version that is stable, and has a minimal set of functions, it is under
continual development, and is not being generally released at this stage. If you
have an interest in using the program now, please contact us.
An explanation of the layout of the space shown in the figures below is given here,
though this is not the only layout of which the program is capable.
The basic interface
Almost all operations are currently performed directly in the main window, which is shown
in Figure 1 (click on all images to see the full-size version). From here, you can interactively
change the dimensionality and alphabet size for the space, scale the drawing with a slider bar,
and choose between the different space representations.
Figure 1. The main Space Explorer window. Here, it shows a 4-dimensional RNA space (i.e.
all possible RNA sequences of length 4 bases). The sequences are colored according to their
hamming distance from the 'master sequence' AAAA.
A binary space is shown in Figure 2, again with points colored according to their Hamming distance
from the master sequence.
Figure 2. This figure shows a 10-dimensional binary space, in which points
are colored according to their distance from the sequence '0000000000'.
Associating data with the space
Showing the space itself is only the first step. This visualization tool is useful for a wide
variety of different purposes, which we are only beginning to explore. One of the most important
will be the display of distributions of sequences in sequence space, in a variety of different
simulations, or the fitness or some other value associated with each sequence. For the first
time, it is now possible to see exactly how a population is clustered in sequence space
in these types of simulations.
Figure 3. An example of some data superimposed on the space; in this case,
one frame of a multi-generation simulation is shown. Both point size and color are being set
here according to the population size of each sequence.
We are already starting to find Space Explorer useful, and have two examples online.
The first concerns the flow in
sequence space of populations of predators and prey in
a simple model; the second,
population movement in a simple NK model.
Last updated 21 August 1997
Copyright © Simon Fraser 1997.