4 ms·

different ligands can bind to the same receptor in different ways and cause different signaling cascades. look up "biased agonism"

by valec 4y ago

tux3 4y ago
It's fascinating how Nature effortlessly creates these leaky abstractions, completely opposite to how we design systems, and makes good use of the constant layering violations where we would see unmaintainable code that we can't wrap our heads around. When we design signaling mechanism, they deal with the layer below while presenting a simpler digital interface to the layer above. Your Ethernet PHY has to care about electrical details, but to the IP stack it's just a stream and sink of bytes. No weird side-band edge cases effects, like if the IP stack could affect DC balance by changing the mix of 1s and 0s it sends out. But those G protein-coupled receptors — or whatever it is — it seems like if you don't want to miss important side-band edge case signals, you really have to model them as this 3D structure that threads through the membrane 7 times, that you can jiggle on one side to change the shape it has on the other side. The receptor-agonist abstraction makes total sense from a human perspective (and it's a good and useful model), but at the end of the day the Wikipedia lists nine different types (partial, super-, inverse, co-, irreversible, biased, ..) of agonists, multiple of which can apparently apply to a single ligand ("can concurrently behave as agonist _and_ antagonists at the same receptor, depending on effector pathways or tissue type"). Nature seems to have absolutely no preference for clean abstractions, and I don't know how to feel about that, but it's fascinating to me. Maybe my horrible code is actually universally optimal under a non-human optimizer =) Or maybe it's just intrinsically hard to make clean abstraction out of chemistry? I wonder if it _would_ be beneficial for a large system like the body to be built out of cleaner (less powerful) abstractions that don't expose all their internal details, but maybe stochastic DNA mutations with natural selection isn't a good enough optimizer to find that place in humanity genomespace? Death is so out of style compared to Adam and SGD.
- stackbutterflow 4y ago
  I think that's called five-billion-years-of-trial-and-error driven development.
- nonrandomstring 4y ago
  > like if the IP stack could affect DC balance by changing the mix of 1s and 0s it sends out. Brilliant! I'm going to steal that for a datacoms lecture to hammer home 'layers'.
- magicalhippo 4y ago
  > like if the IP stack could affect DC balance by changing the mix of 1s and 0s it sends out For those that don't know, avoiding DC bias due to the distribution of 0's and '1 in the transmitted data is a real concern and something that's engineered into many transmission protocols. A common choice is a form of paired disparity code[1] like 8b/10b[2], found in HDMI, DisplayPort and PCIe 1.0/2.0, and variations[3] like 128b/130b for PCIe 3.0 or 128b/132b for USB 3.x. [1]: https://en.wikipedia.org/wiki/Paired_disparity_code https://en.wikipedia.org/wiki/Paired_disparity_code [2], https://en.wikipedia.org/wiki/8b/10b_encoding https://en.wikipedia.org/wiki/8b/10b_encoding [3]: https://en.wikipedia.org/wiki/64b/66b_encoding#Technologies_that_use_128b/1xxb_encoding https://en.wikipedia.org/wiki/64b/66b_encoding#Technologies_...
- kortex 4y ago
  > No weird side-band edge cases effects, like if the IP stack could affect DC balance by changing the mix of 1s and 0s it sends out. As someone who started in chemistry and moved into computers, this made me laugh out loud hard. It's an absurd and absolutely perfect analogy for the weird shit biology does all the time. I think it is intrinsically hard to make clean abstractions out of chemistry. Biochemistry even more so. Chemical synthesis in the lab is all about the loop of: react under ideal conditions, work up, extract, purify, repeat. Biochemistry is just...fancy soup. You get some barrier, especially for really nasty stuff, like peroxisomes, but there is just tons of diffusion. Especially with neurochem, where usually drugs have to be the right lipophilicity to cross the BBB but also specific to targets.
- heavyset_go 4y ago
  I disagree. We build happy paths for our designs, but there are plenty of vectors for attack and modification outside of things like exposed APIs and other "contracts" we bake into systems. You see this all the time with viruses, jailbreaks, virtualization, and hardware hacking. If we had instrumentation that allowed us to poke and prod at microscopic integrated circuits, there would be a lot of attack vectors exposed there, too. After all, our abstractions are built upon implementation details, and if you can play with the implementation itself, those abstractions and contracts are up for grabs when it comes to modifications.

different ligands can bind to the same receptor in different ways and cause different signaling cascades. look up "biased agonism"

by valec 4y ago

different ligands can bind to the same receptor in different ways and cause different signaling cascades. look up "biased agonism"

tux3 4y ago

It's fascinating how Nature effortlessly creates these leaky abstractions, completely opposite to how we design systems, and makes good use of the constant layering violations where we would see unmaintainable code that we can't wrap our heads around. When we design signaling mechanism, they deal with the layer below while presenting a simpler digital interface to the layer above. Your Ethernet PHY has to care about electrical details, but to the IP stack it's just a stream and sink of bytes. No weird side-band edge cases effects, like if the IP stack could affect DC balance by changing the mix of 1s and 0s it sends out. But those G protein-coupled receptors — or whatever it is — it seems like if you don't want to miss important side-band edge case signals, you really have to model them as this 3D structure that threads through the membrane 7 times, that you can jiggle on one side to change the shape it has on the other side. The receptor-agonist abstraction makes total sense from a human perspective (and it's a good and useful model), but at the end of the day the Wikipedia lists nine different types (partial, super-, inverse, co-, irreversible, biased, ..) of agonists, multiple of which can apparently apply to a single ligand ("can concurrently behave as agonist _and_ antagonists at the same receptor, depending on effector pathways or tissue type"). Nature seems to have absolutely no preference for clean abstractions, and I don't know how to feel about that, but it's fascinating to me. Maybe my horrible code is actually universally optimal under a non-human optimizer =) Or maybe it's just intrinsically hard to make clean abstraction out of chemistry? I wonder if it _would_ be beneficial for a large system like the body to be built out of cleaner (less powerful) abstractions that don't expose all their internal details, but maybe stochastic DNA mutations with natural selection isn't a good enough optimizer to find that place in humanity genomespace? Death is so out of style compared to Adam and SGD.

stackbutterflow 4y ago
I think that's called five-billion-years-of-trial-and-error driven development.
nonrandomstring 4y ago
> like if the IP stack could affect DC balance by changing the mix of 1s and 0s it sends out. Brilliant! I'm going to steal that for a datacoms lecture to hammer home 'layers'.
magicalhippo 4y ago
> like if the IP stack could affect DC balance by changing the mix of 1s and 0s it sends out For those that don't know, avoiding DC bias due to the distribution of 0's and '1 in the transmitted data is a real concern and something that's engineered into many transmission protocols. A common choice is a form of paired disparity code[1] like 8b/10b[2], found in HDMI, DisplayPort and PCIe 1.0/2.0, and variations[3] like 128b/130b for PCIe 3.0 or 128b/132b for USB 3.x. [1]: https://en.wikipedia.org/wiki/Paired_disparity_code https://en.wikipedia.org/wiki/Paired_disparity_code [2], https://en.wikipedia.org/wiki/8b/10b_encoding https://en.wikipedia.org/wiki/8b/10b_encoding [3]: https://en.wikipedia.org/wiki/64b/66b_encoding#Technologies_that_use_128b/1xxb_encoding https://en.wikipedia.org/wiki/64b/66b_encoding#Technologies_...
kortex 4y ago
> No weird side-band edge cases effects, like if the IP stack could affect DC balance by changing the mix of 1s and 0s it sends out. As someone who started in chemistry and moved into computers, this made me laugh out loud hard. It's an absurd and absolutely perfect analogy for the weird shit biology does all the time. I think it is intrinsically hard to make clean abstractions out of chemistry. Biochemistry even more so. Chemical synthesis in the lab is all about the loop of: react under ideal conditions, work up, extract, purify, repeat. Biochemistry is just...fancy soup. You get some barrier, especially for really nasty stuff, like peroxisomes, but there is just tons of diffusion. Especially with neurochem, where usually drugs have to be the right lipophilicity to cross the BBB but also specific to targets.
heavyset_go 4y ago
I disagree. We build happy paths for our designs, but there are plenty of vectors for attack and modification outside of things like exposed APIs and other "contracts" we bake into systems. You see this all the time with viruses, jailbreaks, virtualization, and hardware hacking. If we had instrumentation that allowed us to poke and prod at microscopic integrated circuits, there would be a lot of attack vectors exposed there, too. After all, our abstractions are built upon implementation details, and if you can play with the implementation itself, those abstractions and contracts are up for grabs when it comes to modifications.