Re: Thanks for the suggestions - was rendering in 3D

On May 12, 2006, at 10:02 AM, Bushnell, Timothy wrote:

> Thanks to everyone who suggested possible software to view data in  
> 3D.  I’ll be trying several different platforms to see which works  
> best for our applications.
>
> The suggested platforms include (in alphabetical order):
>
> Coulters CXP software for the FC500
>
> Rflowcyt
>
> Weasel: http://www.wehi.edu.au/cytometry/WEASELv2.html
>
> Winlist 3d from Verity House Software (www.vsh.com)
>
> Regards
>
> Tim
Tim,

With all due respect to these solutions, you shouldn't think that  
Mario could sleep at night if anyone could perform high dimensional  
analysis better than FlowJo.  [Disclaimer:  Yes, I live off FlowJo  
sales, and that's a blatantly commercial statement, but it gets  
technical from here on.]
We've played with spatial 3D plots a lot, both our own prototypes and  
others, and they just don't do a very good job of discriminating  
populations.  And its impossible to coherently describe the  
populations you can see.   I like the slice-and-dice approach that  
you get by making a flipbook on X vs. Y in slices along the Z axis  
(it shows up as a Quicktime movie), but that too is very difficult to  
use in a way that's better than two 2D graphs connected by a gate.
If you really want to increase your dimensionality, we're just adding  
a new "Polyvariate Plot" to the Mac version of FlowJo for next week's  
ISAC.  The idea was taken from RFlowCyt.  We've added interface  
refinements to make it more interactive, but like any good R tool,  
it'll astound and confuse you.
http://www.flowjo.com/v8/html/polyvarplot.html

The Polyvariate Plot can model transformations in any number of  
dimensions.  So it will produce a 3D plot, or as many dimensions as  
you want (Shown below in 5D).  And it projects these transformations  
onto a graph window, so you can gate on them.

We're still trying to figure out the applications for this  
visualization, but if you're looking for another dimension as a way  
to differentiate populations, we think this is potentially much more  
powerful than conventional spatial projections.   This is explained  
in the poster P178 at ISAC next week, or at the web page above.

Be forewarned: This is the opposite end of the sizzle/steak  
spectrum.  Most people use 3D graphs to make their PowerPoints look  
spiffy.  These graphs are absolutely impossible to explain in a  
presentation.

Adam

-----------------------

A 3D plot:



A 5D plot: