Hacker News Reader

Comment by udev | Hacker News Reader

The amount of information required to orchestrate such a process seems incredibly high compared to the amount of data stored in the DNA.

My understanding of Information Theory completely breaks down looking at this. It seems impossible that it works.

I have this feeling that we have not yet developed the theories that would make it possible for us to grasp how this truly works.

View on HN

The process is riding on a foundation of chemistry and physics that has a whole lot of intrinsic dynamics. The DNA only has to take a causal role in modulating those existing dynamics. It doesn’t have to orchestrate every minuscule interaction, only coordinate things to the extent that the intrinsic dynamics come together to form something with an emergent complexity. It’s like calling into a complex library with a line or two of python. Yes, the top level is compact, but only because it’s built on top of a base of complexity.

That explained it beautifully.

To add to this: It's like the floating water level valve in a tank. It accurately controls the level of the water in the tank with a very simple mechanism. It does not control where every water molecule goes.

I think I get what you're trying to say.

But it does not really explain certain phases of this process, in the beginning, where cells are dividing exponentially and suddenly a fold appears, and simultaneously many cells, many being far apart, start behaving differently from other cells just next to them.

So whatever chemistry or physics foundation helps achieve that, it has this capacity to be very selective (these cells far apart do it, but these ones next to them don't), i.e. it is not temperature, acidity/base concentration, pressure, or other basic physics, because these phenomena don't have this kind of selectivity.

I think such basic phenomena have more power than it would seem. I’m not a biologist, but from poking at wikipedia:

> Turing correctly predicted a mechanism of morphogenesis, the diffusion of two different chemical signals, one activating and one deactivating growth, to set up patterns of development https://en.wikipedia.org/wiki/Morphogenesis

> cells move to be near other cells of similar adhesive strength in order to maximize the bonding strength between cells and produce a more thermodynamically stable structure https://en.wikipedia.org/wiki/Differential_adhesion_hypothes...

Though it is sort of well understood in embryology. It was quite some times ago I studied it, but the big picture is that each cell starts out as a so-called pluripotent cell, that can turn to every other cell type. The important first step is axes selection (head-tail, left-right). The left-right axes is selected in a really interesting way, cells in the middle have a little flagella that tends to rotate in a specific direction. (If my memory is right, the condition where someones internals are all mirror images of “normal” is caused by some problem with these) Now secreted signal molecules will be concentrated on one side, and this selects the given side.

But the big picture is that each cell has a decision tree, and local signal molecule concentration is the criteria on which they decide. And later each cell has the ability to express yet another signal molecule and so on. If we were to imagine a 2d plane and you could symmetrically place a secretion center of A and B signal in a left and right region respectively, than you can have a new type of cell in the middle if it changes based on “signal A and B is on”.

A complex organism uses the DNA to its full potential. For example, some genes are regulated by a large number of enhancers, which results a combinatorial complexity. That is, you don't need a large number of regulators for your system. Instead, there's just a handful, but they're mixed and matched to create thousands of combinations. Alternative splicing and post-translational modifications contribute to complexity as well, as you can create hundreds of thousands of protein variants with relatively few genes. And not to forget that genome is highly dynamic: there're constant rearrangements, which allow producing complex patterns of gene expression.

Have you watched demoscene videos? I think you're overestimating how much information is needed.

Yes, I watched.

I agree those can be very compact, and looks like a whole world is depicted in 4kb with music!

Demos are based on combining perfect geometrical structures in clever ways (planes, spheres, sinusoids, etc) because they all can be compactly expressed with the type of algebra/arithmetic/geometry that is exactly what our CPUs were built for.

If you watch many demos you tend to see certain symmetries, reflections, and other artifacts that make it all seem repetitive and plain (despite the number of moving objects, effects, etc.)

The process in this cell dividing feels very different to me, but it is hard to put my finger on it.

Watching the initial stage where this amorphous blob of cells just kept multiplying, made me think that the blob of cells is just that - the bioprinter material. The shape is probably set by the printer, i.e. the parent body, and that blob of cells would never form anything meaningful without the parent. This theory would fall apart, though, if such an organism can be replicated in a lab.

"The bee, of course, flies anyways. Because bees don't care what humans think is impossible."

On the contrary, there is alot of literature on this fascinating subject, many books and papers, TL;DR - as a first approximation:

(1) In asexual reproduction, each successful offspring generates about 1 bit of useful information ("entropy")

(2) In sexual reproduction, each generation of offspring from population n can generate on the order of log(n) bits of information

Ref: Information Theory, Inference and Learning Algorithms by David J. C. MacKay

Reply on news.ycombinator.com