In the future, lab mice will live in computer chips, not cages?

https://undark.org/2020/07/02/future-lab-mice-computer-chips/

by MindGods • 6 years ago

58 78 6 years ago

This is, quite frankly, ridiculous. Yes, computational models do have a place in science and drug discovery, but I don't know anyone in the field who seriously thinks they will reduce the number of animals used in medical research.

If anything, you may see something like the Jevons paradox, where when you increase the efficiency of something (in this case, animals used in medical research), usage actually goes up.

theonemind• 6 years ago

I don't think it'll happen tomorrow, but computer technology hasn't finished advancing...nor has our medical knowledge. Perhaps within the next century bits will turn out cheaper, faster, and more effective, perhaps even enabled by quantum computing. Biology seems massively parallel, and we've only just gotten started with parallelism in computing.

montereynack• 6 years ago

That literally makes no sense. How exactly would a quantum computer assist us in simulating problems like drug transport and binding? (quite a few problems are purely empirically solved right now) How exactly is biology massively parallel? I suppose you could argue a cell is just a weird form of SIMD but that is a vast oversimplification. We’ve also been doing parallelism pretty much since computers were invented (and even before that, assembly lines were a thing) it’s just that now we really care cause our one trick of shoving transistors in doesn’t work so hot anymore.

autoexec• 6 years ago

If anything, you may see something like the Jevons paradox,

Thank you for that. I think it might help explain why no matter how large my hard drives get I seem to have no problems filling them as the sizes and number of files seemed to increase to keep me constantly upgrading.

As for the lab mice, if we ever do reach the point where models are enough to replace the need for live animals it does open the door for even the most casual experimentation that might be considered too cruel and inhumane to be performed on living creatures. I'm not sure what kind of useful results could come from that kind of research, but at least then it could be explored without moral considerations or causing trauma to the researchers who would otherwise have to become torturers.

newsbinator• 6 years ago

That's probably what the beings who are simulating us were thinking.

rhn_mk1• 6 years ago

I don't know about the state of current tests, but it doesn't seem to me like the author is necessarily proposing to switch left and right to use computer models instead of a real mice.

Instead, the way I'm reading it, they are recommending to use computer models whenever possible, and only resort to live animal testing when the limits of the modelling are reached.

chrisco2553• 6 years ago

It's a waste of time. There isn't a model that can sufficiently simulate the chaotic behavior of biological cellular systems and the complex interactions with potential chemical treatments. Models are Mickey mouse representations of reality.

bedobi• 6 years ago

Yeah, no.

Programs and algorithms can only encode models (which are themselves imperfect) of what we already know. There's countless things we know that we don't know, and yet more that we don't know that we don't know.

As bad as animal (and human) studies and trials can be, the beauty of them is we don't really have to know how or why things work, we can just do the thing and if the results are good, we have a new treatment. Lots of super common drugs, even paracetamol, are like this.

Of course, knowing and accurately modelling the action is great, but it's not necessary.

skwb2• 6 years ago

This is why I think the quote "All models are wrong, but some are useful" is so meaningful. Lots of bright eyed comp sci majors come into statstical genomics and bioinfomatics PhD programs every year, and all learn the hard lesson that wet lab still exists for a reason.

Not understanding your underlying conceptual model (be it biology here, or maybe healthcare, or natural resource exploration) will lead a lot of poor assumptions, limiting the impact of all the hard work.

chrisco2553• 6 years ago

I would argue we can't even model those things. Non-linear, dynamic systems are not predictable beyond a certain point even if you have a perfect model. Uncertainty with models of dynamic systems like biological ones, increases exponentially as a function of time. And they are highly sensitive to initial conditions, so if your input is off from reality by even the tiniest amount, reality will veer off dramatically from your expected results, even if you have the perfect model. Chaos theory is a thing. We can use models to learn about systems and how they might work, or how we might think they work, but we can't use them to model reality effectively enough to determine treatments.

chr14• 6 years ago

While individual trajectories of non-linear systems are not predictable, they all end up on an attractor and usually all we are interested in, is the attractor.

That is when performing a test on a mouse we are not measuring what a particular mouse would do, but what the mice do on average. So as long as model behaves like a mouse it can be useful, even though it would not behave like any particular mouse.

Also it's important to remember that the question is not about all or nothing, when we get a model most likely instead of 100 experiments on real mice we will be doing millions on virtual mice, and then 10 on real mice to verify the results.

gridlockd2• 6 years ago

As bad as animal (and human) studies and trials can be, the beauty of them is we don't really have to know how or why things work, we can just do the thing and if the results are good, we have a new treatment.

If you didn't have at least a broad idea of why something might work, you would most likely never have "done the thing". You need to narrow the search space. Simulations can help here.

knowmylimits• 6 years ago

ML models sometimes cannot be explained by humans and contains information we do not know. But they can be contained in a program or algo.

theptip• 6 years ago

Programs and algorithms can only encode models (which are themselves imperfect) of what we already know.

The Game Of Life seems to contradict this claim. Also worth considering, computational physics and weather/climate simulations.

Scientists of all sorts already do learn new insights by creating simulated models of systems. The simulated systems will continue to get more accurate.

RcouF1uZ4gsC• 6 years ago

Doing drug development without animal testing is like trying to modify a huge, messy, legacy code base without integration tests.

Our computer models are like unit tests. They are relatively self-contained and quick to run. However, they cannot test all the interactions and stuff we do t know about, so we have animal testing as a kind of integration test to make sure stuff works like expected.

And a phase 3 clinical trial is like pre-prod. When you think everything works, you expose test it in the real world with very careful monitoring and plan to keep it from going out into prod if problems are found.

foobar_6• 6 years ago

Its pointless to use animals in research, it makes no logical sense. Even if the research is successful after 10 years of animal cruelty ... you still have to test it on humans.

sk0g2• 6 years ago

What about all the chemicals that never get to human testing due to harmful effects coming up in animal testing? Is that not a net positive from a human perspective?

foobar_6• 6 years ago

Are there absolutely no alternatives to test that ?

vixen99• 6 years ago

Sure - though plenty of chemicals hurt animals but don't hurt us. One small example: theobromine and caffeine (in chocolate) kill cats and other animals that lack the ability to metabolize these alkaloids rapidly enough.

Oddly enough, our own cat loves chocolate and (following a near miss & visit to the vet) now appears to resent having had that avenue of pleasure permanently blocked.

adrianN• 6 years ago

Given the current state of our nematode simulation tech (http://openworm.org/), "the future" might take a while to arrive.

causality0• 6 years ago

This is less of a "we will replace natural meat with lab-grown meat" claim and more of a "we will eventually be able to travel to the moon using only a megajoule of energy" claim. It would take a vast change in our basic understanding of physics to effect such an innovation. The crux of the matter is that mouse bodies don't simulate mouse bodies, they are mouse bodies. The idea you could have something the size of a mouse body that could fully simulate a mouse body is ridiculous. Biology is computationally efficient.

The University of Nevada used a 360 TFLOP supercomputer to simulate a mouse brain. They managed to simulate the neurons firing once every ten seconds. In reality neurons fire an average of 200 times per second. Even if their model was fully accurate, that means you need 720PFLOPS to simulate just the brain of a single mouse.

Barring computers based on completely different principles or a revolution in the way we understand physics, a installation at a single facility will never be able to fully simulate hundreds of experimental mice.

bserge• 6 years ago

It's absolutely amazing to me how efficient and powerful brains are. Just the simple action of say, moving from one side of a yard to another involves realtime sensor information, motor control and feedback, threat assessment and decision making, previous experience recall and more. In mere miliseconds even in mice.

It gets insanely more complicated for humans. If you decide to drill a hole in a piece of wood, your brain will process a ridiculous amount of information just for that single purpose, plus it's always running primitive functions like any other animal. Realtime.

I don't see any computer doing that anytime soon.

cellis• 6 years ago

Who here has watched as much Black Mirror/Westworld as I have and know how this ends? My question is, if a robot is engineered well enough to replicate the "real" thing, then when you test on it, aren't you just shifting the torture from "living" to another thing yet to be granted the rights of the living? If you answer "no", then you have to admit that the model is incomplete and it does not in fact feel pain and is thus in an inadequate replica. Also agree with other comments that this is and will remain computationally infeasible at least until production quantum computers arrive.

ars3• 6 years ago

You can ask the reverse question - if the Robot version of a mouse is sufficiently similar to a mouse, maybe the mouse also doesn't actually feel pain, and is really a kind of biological robot.

Maybe so are you.

Dualism holds that living entities are special, and can never be replicated artificially. So your question can actually be turned into dualism vs monism (which is not a question with a settled answer).

dTal• 6 years ago

This comment doesn't make any sense to me. Why would the existence of a robot copy imply a lack of pain? How can you argue that whether humans feel pain is an unsolved question? Shall we experiment on you instead?

FeepingCreature2• 6 years ago

This would run into the "here is a hand", or rather "here is a fist" objection.[1]

Pain is the ground reality that is to be explained. If you manage to explain away pain rather than explain it, you haven't actually answered the question at all. If your answer at the end is "pain is not real", this doesn't prove that pain is not real, it proves that your premises were wrong.

[1] https://en.wikipedia.org/wiki/Here_is_one_hand

numpad0• 6 years ago

Easiest way I can come up to go with it is to admit we sometimes exploit life for the greater good, and also swear to yourself that you do your best to keep it as natural, efficient and minimal as possible. Otherwise you don’t get to even grow and eat single grain of wheat do you...

cosmodisk2• 6 years ago

Realistically this unlikely to be a problem in this century, however you are raising some valid points nonetheless. A lot of,if not most,capabilities animals, including humans, have are related to one thing: survival. Ability to move rocks, climb trees,or build sophisticated structures boils down to body's need to get food and preserve itself from destruction. These aspects are often overlooked when modelling AI and the whole focus is on pure logic. I sometimes think that one day we'll end up with universal AI and the first thing it will ask will be sometime like: "why should I do it?".

ejolto• 6 years ago

If you don't do it our funding will be cut and we'll have to power you down.

totetsu• 6 years ago

Ian M Banks explored this in the culture series. When doing speculative simulation of how another civilization will react to situation, what do you do with all those little simulated individuals when you've got the data you need. If you simulated them with enough detail, isn't turning the simulation the same as genocide?

netcan• 6 years ago

I'm not sure these questions can be answered philosophically. After thousands of years, philosophy can't even agree on the the morality of things that already do exist.

Science fiction and moral philosophy depict "model organisms" as physical replicas of organic ones for a reason. Our moral ponderings are driven by moral feelings, and they need to trigger feelings for the reasoning to be interesting. The neocortex court only debates when limbic prosecutors press charges.

Ultimately, the task here is software that answer questions, not software that is a mouse. If software those two converge, we have a moral dilemma. Seems unlikely though.

chr14• 6 years ago

If you torture someone then use a time machine to restore him to the initial state, did you do anything bad?

You can't do this to anyone real but you can easily save and then restore the state of the simulation at any point.

Torturing someone in a simulation is not any different than imagining someone being tortured, or performing lots of arithmetic operations.

So the only logical thing is to admit that torturing/killing/hunting a species to extinction is bad not because of pain, but because of stealing computational resources belonging to someone else/some other sophisticated computational process.

wizzwizz44• 6 years ago

Torturing someone in a simulation is not any different than imagining someone being tortured,

Isn't it? Our imaginations are predictive, not simulative.

Digit-Al• 6 years ago

I think you are missing something obvious here. Pain does not exist in any objective sense, it is a construct created by the brain in response to certain stimuli as a survival mechanism. If you were creating a simulation of a human body to use in medical experiments then surely you would not give it a brain that functioned in any meaningful sense.

I would not say that a body incapable of feeling pain is an "inadequate replica". A human or animal under anaesthetic does not feel pain - are they inadequate. You can certainly perform surgery on an unconscious person, and you can administer medications to them.

[edit] I should have added that I am totally in agreement that we are not even close to being able to simulate a human body with any degree of accuracy.

nunja• 6 years ago

Maybe pain is a part of the treatment as any body/antibody response. It is a push notification mechanism designed in a million of years of evolution for a reason. But I agree that it might be deprecated soon.

byteface• 6 years ago

I worked with a guys who said 'one of the reasons' he quit medical research was because of the amount of mice he had to kill on a daily basis. He said, "the quick way is to snap their necks with the back of your pencil." as he demonstrated on the desk the technique of how to hold the pencil a certain way.

Gatsky• 6 years ago

This isn't the right point to make. There are a lot of unnecessary animal experiments because the way in which experiments are planned is very unprincipled. in silico approaches have great potential to improve the efficiency of animal experiments.

lordnacho• 6 years ago

Sounds like map-vs-territory confusion to me. A model of a mouse in a computer isn't a mouse, the point of doing the experiment is to uncover whether what you thought the model predicted was in fact what happened, because it's possible you didn't incorporate some aspect of the real mouse.

Simulations are useful for making calculations. Say I'm convinced that buildings work a certain way and I have a finite element model of one. I don't know what the predictions are for a new building of a certain shape, so I shove it in a computer to tell me. The computer tells me what I should predict assuming that my finite elements ideas were correct. What it won't tell me is whether there's some aspect of that model that doesn't correspond to what the building is actually gonna be made of.

ganzuul• 6 years ago

This is apparently different from the lab-on-a-chip research which seems to seek to miniaturize organs into a microfluidic environment. I'm not actually sure if they are using organ cells on a scaffold or if they can discard the scaffold for microfluidics and cells alone even after reading a few papers on the subject.

Personally I think living 2D animals in a petri dish would be metal as fuck.

jfoutz• 6 years ago

A friend of mine works on simulating mouse organs. We had some pretty great talks over beer about modeling and cross model validation. Eventually he'd rant about all models are wrong but some models are useful.

If I remember correctly, they had some success in rejecting animal studies. Things like, we don't know what this will do overall, but we're pretty sure the kidneys or liver will fail like X, so don't bother with the wet lab. Or possibly try the wet lab at small scale so they can validate the model.

We're far far away from full simulation. But we know a whole lot about mice. I internalized those conversations as, my friend writes type checkers for drugs against mouse hardware.

teekert• 6 years ago

This completely bypasses the revolution that will take place long before this: Organs on a chip. With the development of organoids [0] we made the first step. Things like the lung-on-a-chip [1] are steps towards a sort of mini body that mimics many functions and simulates the body not only in the molecular biological way but in the very important physical way (which, it turns out can not be decoupled from the molecular biology).

Why would we put mice in computer chips if we can have pseudo human bodies in the real world?

[0] https://www.youtube.com/watch?v=2SG5ivm6jkw

[1] https://wyss.harvard.edu/media-post/lung-on-a-chip/

gridlockd2• 6 years ago

Why would we put mice in computer chips if we can have pseudo human bodies in the real world?

Because a lot of the existing data was gathered on mice.

S0und• 6 years ago

Why do they want to model mice, when they the whole medical industry don't know shit about the mice they use for experiments?

They use mice to experiment with, BUT through breeding the current lab mice are more resilient then humans. Which can imply that certain drugs getting passed for human testing because they did well with mice. Today's lab mice are more resilient, but the same drug can damage humans.

Bret Weinstein talks about this on JRE: https://www.youtube.com/watch?v=pRCzZp1J0v0&t=9597s

paulchap• 6 years ago

I guess now would be a good time to recommend 'Permutation City' by Greg Egan... Great book!

olodus• 6 years ago

One can look at it like how us developers look at testing. You have multiple steps on the way from code to delivery. There are unit test, more realistic test frameworks, test environments and then finally tests at a customer / usage by a customer.

In the same vain I could see (and even hope) some mice tests could be decreased or removed as better models and simulations gets created. That would be quicker and more cost efficient for everyone. But in the end you want a real world test before giving it to a human. And maybe we could find better things than mice for that as well, but right now I think that is the best we have got. I am not a medical researcher though so I don't know if there are alternatives out there.

zalkota• 6 years ago

The brain is not an algorithm and can’t be described as such.

rjmo• 6 years ago

As a fellow animal researcher: this article is pretty much devoid of informed or carefully considered thought about the issue.

Many commenters have noted that biological systems are mind-numbingly complex, and we're not even close modeling these systems in a way that's useful for testing novel hypotheses of biology researchers.

The field in which I work, neuroscience, frequently does employ computational modeling, but this modeling is primarily used to provide a mechanistic link between inputs and outputs observed in an in vivo experiment. Present a sensory signal to an animal, observe something about neural dynamics in a brain region, THEN use simplified biologically-inspired elements to provide a plausible mechanism between the two. Even the state-of-the-art in silico model is always child's play compared to the real thing. We definitely don’t know enough about that real thing to build a model that would circumvent its need. This author also works in a neuroscience lab – they should know better.

In the future, lab mice will live in computer chips, not cages?

Related Stories