On a second thought, it seems it could be used for 3D shapes in genetic algorithms, with each English letter representing one neighboring voxel. If it hods true that similar DNA results in a similar shape, it could work very well.
It doesnt seem like a lot of thought has gone into making the colour / drawing choice actually visually significant. Eg. given blue is much harder to see here than red, are we over-estimating similarity because there is a lot of red?
I suspect they need to talk to some visualization & human vision people to make this compelling.
Like Pepe said below, each letter spawns in a specific diredtion, so A, or red, always goes up, and you can check that in the beginning, where there are only few lines. so the final shape is more important than color. the color is basically irrelevant, except for the fact that maybe you can see which of A, C, G, T appears more times in that dna sequence.
Can someone clarify this for me? From watching their video, it seems like all they've done is build a tool that visualizes DNA? I don't understand how the tool assists with visual differences in DNA. I expected maybe a single DNA sequence in one color with the differences highlighted in a contrasting color. Maybe the video just didn't do a good job of explaining things.
Although the 3D plots are neat, they kind of feel like 3D barcharts: they add very little over a more straightforward representation (like a heatmap instead).
9 comments
[ 5.0 ms ] story [ 31.4 ms ] threadI suspect they need to talk to some visualization & human vision people to make this compelling.
Couldn't resist trying something similar with 2D canvas:
https://output.jsbin.com/zorepe
It will definitely look better in genuine 3D, but that's a bit more painful. Totally doable, though.
http://grondilu.github.io/dna.html
http://sandwalk.blogspot.com/2012/01/whats-difference-betwee...
Although the 3D plots are neat, they kind of feel like 3D barcharts: they add very little over a more straightforward representation (like a heatmap instead).
There might be more recent approaches like the one mentioned by aaronmck which is from 2012.