Understanding Hormones and Data Transfer

Cells can do distributed computation via hormone diffusion, in which the concentration of a hormone at a cell is a proxy for distance in 3D space from a source emitting it

Every cell is basically a 3D shape with these 3D binding sites. A molecule that has a particular 3D structure can fit into this site like lego. We call the molecules that do this hormones. They are signalling molecules - meaning their structure is their only function - they are not little machines like enzymes/proteins that build things. They are just unique shapes that can be used to transmit 1 bit of information. For example, insulin is a hormone. It has a particular shape that can only bind to a insulin receptor on a cell. When it travels along the blood stream, and encounters a cell, it might bind to the site. When that site is bound, the cell now has a value of 1 in that area. When it doesn’t, it is 0. It can use that value to conditionally start certain internal processes, like the uptake of glucose.

Cells have hundreds of thousands of receptors. For example, a human blood cell has around 100k insulin receptors. This means it can receive 100,000 bits of signal for the single “insulin” value, since a receptor can either have a molecule bound (1) or it can be free (0). They call this a concentration, and over time, a hormonal gradient.

In a sense, abstractly you can think of the concentration as a single value

ie. a cell has 100k insulin receptors, each a 1 or a 0. log2(100_000)=16.6 bits uint16 insulinConcentration;

The geometry of a vascular network is interesting. Imagine a leaf. You have the tissue of the leaf, and then you have the veins running down it. The veins are the transport system for the molecules/hormones.

When a cell at the part of the leaf closest to the plant emits hormones, it travels along the vascular network. Each hormone is a molecule. When it travels down the vascular system, it may bump into a cell that binds it. After that point, it is taken up and no longer travels. The binding is not guaranteed - these are little particles in 3D space.

This emitter of hormones might emit 1M hormone molecules. Imagine each cell in the network only has 100k receptors. That means those 1M hormone molecules might “fill up” the receptors of 10 cells at best. Which cells will get filled up first? The ones closest to the emitter. The ones furthest away have no chance, since those hormone molecules were already taken up.

What’s interesting to me - and this is what I learnt from building the Origami paper yesterday (https://x.com/liamzebedee/status/2050459472955646045) - is that this hormonal gradient mechanism is sufficient enough for cells to coordinate distributed computation

The hormone concentration + a vascular network means that a hormonal concentration can act as a distance metric from an emitting cell, where the concentration’s precision/quantization is determined by number of the cell’s receptors for that hormone.

It is more easily understandable if I show you code, which I don’t have, because I only just figured out how this works ;) hahahhaha

The example I think is pertinent is “grow a cylinder of fixed height” by writing code that runs the same on every cell, where each cell has a small scratch space for internal state, and the ability to send/receive hormones

The code might look like:

divide()

But that grows an infinitely expanding 3d thing. You want it to only grow to a maximum height.

How do you measure height, when every cell runs the same program? You could emit a hormone at the “top cell” in a concentration that effectively “runs out” by the time it reaches the bottom cell of the cylinder.

I’m still working on figuring what the code looks like out. But I sort of have the shape in my head now.

Insulin isn’t really one of the hormones used for signalling distance. I think those are morphogens but I’ve still got to properly look at them. That and all of this is an approximate mental model - but roughly speaking it’s all very new and fundamentally interesting to me as a way of distributed computation.