Re:A Lizard with Scales That Behave Like a Cellular Automaton
by jfarlow on Wednesday April 12, 2017 @10:26PM
(#14103640)
Turing's paper [1] is pretty impressive. Well before the mechanics of molecular biology were understood, Turing was able to reason out how you could use a 'few lines of (reusable) (biological) code' to produce patterns as sophisticated (and seemingly unrelated) as zebra stripes, leopard spots, and even the shapes of tissues, the number of digits on a hand, and the morphology of even a whole organism. And he was mostly correct once we came to be able to actually describe the physical systems biology uses for patterning.
One example of a generalizable code for patterning by 'reaction diffusion':
- Set up a 'node' that has an inverted feedback loop for producing a 'morphogen' (produce less when lots, lots when less).
- Let the morphogen diffuse through a field.
- Digitize the morphogen concentration by some thresholded sensor.
- Use an orthogonal 'clock' mechanism to freeze the field in different states digitized above.
Generally though the simulation of biology runs into just way way way too many variables to be 'nicely' modeled. Part of the cool part is that when you have a routine that generates a spatial pattern established, tweaking parameters here and there create massive variation while still maintaining a coherent 'pattern'. A bit of chaos gives each run a fingerprint - literally.
One of the other biological patterning mechanisms I like is the Notch pathway, relying on similar principles. [2]
There are fewer variables in a chemical system, but similar reaction-diffusion related principles can be found in chemistry - which can often be 'pretty'. See the Belousov-Zhabotinsky reaction [3].
As an aside, I like that they reference the paper that describes now chameleons change color; "chameleons shift colour through active tuning of a lattice of guanine nanocrystals within a superficial thick layer of dermal iridophores." [4]
One example of a generalizable code for patterning by 'reaction diffusion':
- Set up a 'node' that has an inverted feedback loop for producing a 'morphogen' (produce less when lots, lots when less).
- Let the morphogen diffuse through a field.
- Digitize the morphogen concentration by some thresholded sensor.
- Use an orthogonal 'clock' mechanism to freeze the field in different states digitized above.
Generally though the simulation of biology runs into just way way way too many variables to be 'nicely' modeled. Part of the cool part is that when you have a routine that generates a spatial pattern established, tweaking parameters here and there create massive variation while still maintaining a coherent 'pattern'. A bit of chaos gives each run a fingerprint - literally.
One of the other biological patterning mechanisms I like is the Notch pathway, relying on similar principles. [2]
There are fewer variables in a chemical system, but similar reaction-diffusion related principles can be found in chemistry - which can often be 'pretty'. See the Belousov-Zhabotinsky reaction [3].
As an aside, I like that they reference the paper that describes now chameleons change color; "chameleons shift colour through active tuning of a lattice of guanine nanocrystals within a superficial thick layer of dermal iridophores." [4]
[0] (actual research referenced) http://www.nature.com/nature/journal/v544/n7649/full/nature2...
[1] (PDF! Turing's 1952, The Chemical Basis of Morphogenesis) http://dna.caltech.edu/courses/cs191/paperscs191/turing.pdf
[2] (Notch Patterning) http://www.cell.com/trends/cell-biology/fulltext/S0962-8924(...
[3] (B-Z Reaction) https://www.youtube.com/watch?v=3JAqrRnKFHo
[4] (chameleon's color via guanine nanocrystals) https://www.nature.com/articles/ncomms7368