Can a bacteria be described by physics?

[Borek]

if additional physics would be required.

Can you elaborate?

As I explained earlier, we know all interactions, and we know how to describe them. There are no other interactions - so simulated system should behave just like the real one. And I don't mean we can reproduce exact case and situation, more like we should be able to simulate systems that behaves the way we expect it.

Slightly off topic:

Several years ago I wrote an artificial life simulation of bitozoa. All bitozoa do is they eat, move and reproduce (and evolve, but that's not important here). However, how they reproduce depends on the amount of food. There are two kinds of bitozoa - carnivores and herbivores (there are also plants, again, not important for what I am aiming at). And while I never attempted to reproduce population dynamics, quite often their population oscillate, as if I was solving Lotka–Volterra equation. I don't have a more convincing population history plot here, but I just started the program and got this:

there are four nice oscillations visible (system is not stable, as animals evolve, so they can become better at hunting or avoiding danger).

This is not a simulation of bacteria, however some properties of the system emerge automatically, not because they were programmed, but because that's the way system behaves.

 

[nonequilibrium]

As I explained earlier, we know all interactions, and we know how to describe them. There are no other interactions - so simulated system should behave just like the real one. And I don't mean we can reproduce exact case and situation, more like we should be able to simulate systems that behaves the way we expect it.

But isn't this a circular reasoning? I agree with your statement that given the premisse 'we know all the interactions and how they work' the rest follows, but it's exactly the premisse I don't see as evident. (and I said "circular" because to prove we know everything that goes on (and how), you have to assume we know how everything goes on (and how)) Of course I'm not saying you may not have a good reason to be convinced of the fact we're aware of all the processes, but I don't think you've given any back-up for that? It seems to me the only way to be convinced of that is to actually simulate it using the known and understood processes and see if they suffice. But perhaps I'm overseeing some other verification process.

Note that I'm not saying in advance, with any prejudice, "we definitely don't know all the processes for a bacterium", I'm just stating my ignorance on the subject.
As for the more troubling matter of thought and (self)consciousness, I would say that I do lean more to one side than the other: that we don't know all the fundamental laws just yet, mainly because thought seems to demand a whole new concept that we can't even properly define yet, and I don't know how any other case in the history of science where you had a concept you couldn't even define with your previous concepts yet that new concept followed out of the old concepts nonetheless. Of course this is an opinion/feeling.

 

[Andy Resnick]

Can you elaborate?

As I explained earlier, we know all interactions, and we know how to describe them.

A good quote to think about is "More is different".

http://robotics.cs.tamu.edu/dshell/cs689/papers/anderson72more_is_different.pdf

 

[nonequilibrium]

A good quote to think about is "More is different".

http://robotics.cs.tamu.edu/dshell/cs689/papers/anderson72more_is_different.pdf

I don't find that to be a very clear article; I quote

The main fallacy in this kind of thinking is that the reductionist hypothesis does not by any means imply a "constructionist" one: The ability to reduce everything to simple fundamental laws does not imply the ability to start from those laws and reconstruct the universe.

I would think it does, given the proper computing power, I can't conceive of any reason why not, and the author does not convince me of his point (does he try?).

Anyway, it does not seem like that discussion is what this topic is mainly about: that article is about "okay even if we have all the fundamental laws, can we make/simulate a bacterium" (mixing it with my original question), while I was more concerned with "do we have all the fundamental laws relevant for the bacterium?", taking it as a given that having those laws is enough for also simulating it (again, ignoring technicalities concerning computer power, although of course, for a definite answer, it seems we'd have to check with the computer, so the technicalities aren't completely besides the point, but anyway they're not the point itself)

 

[my_wan]

I don't find that to be a very clear article; I quote

The main fallacy in this kind of thinking is that the reductionist hypothesis does not by any means imply a "constructionist" one: The ability to reduce everything to simple fundamental laws does not imply the ability to start from those laws and reconstruct the universe.

I would think it does, given the proper computing power, I can't conceive of any reason why not, and the author does not convince me of his point (does he try?).

Anyway, it does not seem like that discussion is what this topic is mainly about: that article is about "okay even if we have all the fundamental laws, can we make/simulate a bacterium" (mixing it with my original question), while I was more concerned with "do we have all the fundamental laws relevant for the bacterium?", taking it as a given that having those laws is enough for also simulating it (again, ignoring technicalities concerning computer power, although of course, for a definite answer, it seems we'd have to check with the computer, so the technicalities aren't completely besides the point, but anyway they're not the point itself)

Take a very simple system like a double pendulum, i.e., a pendulum with a hinge in the middle. This system only has two moving parts with laws of motion that are very well understood. Yet we cannot predict how the thing is going to react from any given motion. We understand why we cannot. Now if we consider a system with billions of parts, rather than two, what do you you really expect? No amount of computing power will EVER fix it. Though we can predict those billion parts more accurately than two in some ways.

The range of systems, which even when we fully model on a computer we cannot predict what the computer is going to do from one run to the next with the same data, is huge. You cannot say a programmer did not understand the rules they put in a program. So when you say about predictability you "can't conceive of any reason why not" a whole history and areas of science are being ignored by this statement. Chaos theory, for one, is predicated on these issues. We already understand why we have such limits. Though we can improve those limits we can never make them go away even if we knew every fundamental law in the Universe.

 

[Borek]

Take a very simple system like a double pendulum, i.e., a pendulum with a hinge in the middle. This system only has two moving parts with laws of motion that are very well understood. Yet we cannot predict how the thing is going to react from any given motion.

If you treat the simulation of a pendulum like a prediction, you have a problem. But if you treat this simulation like just another one pendulum experiment, it behaves just like it should. You can't repeat the double pendulum experiment getting the same results twice, why not treat a simulated system as the same experiment run for a third time? Different results? That's what we expected.

Same about bacteria - simulation doesn't have to be about predicting, or perfect recreation of a particular organism, I agree that's impossible. But once you have a running simulation you don't have to be able to point to a particular living organism and say "that's my model". It is enough that your simulated organism behaves like other bacteria - no two are identical, simulated one will be just another one.

Sorry, I am in hurry, I would love to read the paper Andy linked to, but I am leaving for a month in two hours.

 

[DaveC426913]

If you treat the simulation of a pendulum like a prediction, you have a problem. But if you treat this simulation like just another one pendulum experiment, it behaves just like it should. You can't repeat the double pendulum experiment getting the same results twice, why not treat a simulated system as the same experiment run for a third time? Different results? That's what we expected.

Same about bacteria - simulation doesn't have to be about predicting, or perfect recreation of a particular organism, I agree that's impossible. But once you have a running simulation you don't have to be able to point to a particular living organism and say "that's my model". It is enough that your simulated organism behaves like other bacteria - no two are identical, simulated one will be just another one.

Sorry, I am in hurry, I would love to read the paper Andy linked to, but I am leaving for a month in two hours.

Thank you. That's what I was trying to say. Simulation =/= prediction.

 

[Andy Resnick]

Same about bacteria - simulation doesn't have to be about predicting, or perfect recreation of a particular organism, I agree that's impossible. But once you have a running simulation you don't have to be able to point to a particular living organism and say "that's my model". It is enough that your simulated organism behaves like other bacteria - no two are identical, simulated one will be just another one.

Thank you. That's what I was trying to say. Simulation =/= prediction.

But then what's the value of the simulation? All science has value because it leads to *predictions*. It's not sufficient to simply say "here's a simulation that agrees with observations", because that doesn't lead to any new understanding- only a confirmation that a sufficient number of parameters can be blindly tweaked to replicate observed experiments.

It's not required to predict the moment-to-moment state of a bacterium (yet), but surely it would be of value to predict the action of a particular antibiotic.

 

[Ryan_m_b]

But then what's the value of the simulation? All science has value because it leads to *predictions*. It's not sufficient to simply say "here's a simulation that agrees with observations", because that doesn't lead to any new understanding- only a confirmation that a sufficient number of parameters can be blindly tweaked to replicate observed experiments.

It's not required to predict the moment-to-moment state of a bacterium (yet), but surely it would be of value to predict the action of a particular antibiotic.

Indeed, to make the simulation we would have to know all the processes of the bacteria in question anyway. What would be interesting is altering the simulation i.e. increasing the concentration of specific chemicals in the environment.

 

[Andy Resnick]

Indeed, to make the simulation we would have to know all the processes of the bacteria in question anyway. What would be interesting is altering the simulation i.e. increasing the concentration of specific chemicals in the environment.

Exactly. We can't yet predict the structure of a single protein given the amino acid sequence- we have no business extrapolating out the 10^10-10^20 increase in complexity needed for a single cell.

 

[Pythagorean]

A caveat though, to the OP. Describing things in physics doesn't require (in fact, tends to avoid) modeling things as "every single atom". We parcel things into their functional parts and model those. We have no interest in what the atoms are doing in a cannonball simulation.

Likewise, we might find a more appropriate modeling scheme for bacteria if we break it down to functional molecules, lipid membranes, whole processes, etc. We don't want to over-reduce the problem or we get bogged down with unnecessary computation.

Unfortunately, this discussion got derailed quite early- mostly becasue nobody supplied any quantitative data to support their assertions.

Specifically, let's consider a MD simulation of a single E. Coli bacteria. This organism has an approximate mass of 1 pg, of which 70% is water. This leads to an estimate of 2*10^10 atoms of water, 6*10^7 ions, 3*10^6 proteins, and 2*10^7 lipids (Phillips et. al. "Physical biology of the Cell", Garland Sci.). Breaking down the proteins and lipids leads to an estimate of 10^10 atoms in the proteins and 1*10^9 atoms in lipids, for a total of about 3.5*10^10 atoms total, just in the bacteria.

The current world record for an MD simulation is 1.1*10^10 atoms

But this isn't really the question that's being asked; The question is not really one of quantification and instead one of qualification. The question isn't whether we have the computing power. It's if we could, would it work; the question is in the title; "Can a bacteria be described by physics". And the point we were making (at least, in my own case) was that there's still lots of intermediary issues (i.e. the complexity).

In other words, we have lots of theories in biology about how different processes work. Eventually, every once in a while, we discover the means of the process. Never, so far, have the known physical laws been violated in this way. It's always continued to come down to the same physics; only the biology (which permutations of physics nature actually uses) changes. So we say organisms utilize QM (plants) or they don't (the brain), but in either case, physics is not changed or violated.

So we are left with wondering if the laws of physics *in and of themselves* are currently sufficient to "explain" a bacterium, or if additional physics would be required.

So far, we haven't needed new physics. Just deeper understanding of what physics are actually taking place in life forms. Of course, some of us think a paradigm will be required to actually explain a bacteria (myself included) but that's magical thinking. We continue to make progress every day with the current physics.

 

[epenguin]

Protein folding - when I last heard (admittedly something like 10 y ago ) they held annual competitions for who could best predict the 3-D structure from the amino-acid sequence. One guy was doing it qualitatively from physico-chemical knowledge about hydrogen bonds and all the other types of interaction and was doing sometimes better than the heavily computational guys. Suggesting understanding is not the same thing as computation, chemistry is not just computational physics of molecules.

Also the study of protein folding is one speciality. When you simulate a bacterial metabolism would you try to put the protein folding simulation into the whole-bacteria simulation? That would be a rather stupid way of doing things. Whether you do or do not completely understand protein folding, if you just put in e.g. their rate of formation, their catalytic parameters (substrate affinity, catalytic rate constants) even if they had been only experimentally estimated, would you have fundamentally lost anything by that? I you don't fundamentally understand every molecular interaction in minerals that determines their mechanical and thermal properties does that make Earth science a special mystery necessarily in need of New Principles?

Large branches of science handle things with phenomenological laws. When we study electrical circuits - thousands of questions on this site - we don't worry too much abot the various laws being only approximate and the parameters, resistivities etc. not simply predictable. Even though, sure, someone also has to chase up those.

Quantitative mathematical and computational study of metabolism is normal science that has been going on for quite a long time.

 

[Ryan_m_b]

Also the study of protein folding is one speciality. When you simulate a bacterial metabolism would you try to put the protein folding simulation into the whole-bacteria simulation? That would be a rather stupid way of doing things. Whether you do or do not completely understand protein folding, if you just put in e.g. their rate of formation, their catalytic parameters (substrate affinity, catalytic rate constants) even if they had been only experimentally estimated, would you have fundamentally lost anything by that? I you don't fundamentally understand every molecular interaction in minerals that determines their mechanical and thermal properties does that make Earth science a special mystery necessarily in need of New Principles?

You might get an appropriate simulation from simulating the things you've mentioned but it won't be useful. If you want to use a simulation to test how an antibiotic works for instance it's vital that you understand the molecular processes.

Fundamentally biology works on the molecular scale, if you want to get meaningful answers of such complex systems you will need to simulate it on that level.

 

[Andy Resnick]

You guys raise excellent points-too many to explicitly quote :). You also raise some interesting points about what a model is used for, what 'understanding' means, etc.

First, I don't think anyone would seriously argue that bacteria, or any organism, is "outside" of our current state of physical understanding- surely, bacteria obey the second law of thermodynamics, for example.

But some context is in order- considering the level of conceptual abstraction used in physical modeling, biology is positively pre-Newtonian. That is, there is *no* abstract concept that can be applied to biological systems in the way "Force" is used to conceptualize dynamic physical systems. Biology has yet to have it's Newton.

"Modeling" in biology currently means something very different than 'modeling' in Physics, and some of the discussion has neglected this important distinction. While there are some quantitative models in biology- the canonical example in my mind is Guyton's model of cardiac regulation:

http://ajpregu.physiology.org/content/287/5/R1009/F2.expansion.html

The majority of 'models' in biology refer to (simpler) organisms that display a trait or behavior that is analogous to human physiology: a mouse model of renal disease, for example. E. Coli is a model organism- it is used as a model (of limited scope) for more complicated systems.

Now, thread title aside, the OP specifically asked if we could model each individual atom in a bacterium and thence construct a quantitative, predictive *computational* model. Again, my answer is that we should not make any claims regarding this possibility, given the current primitive state of MD simulations.

In the meantime, we must use "coarse grained" computational, experimental, or theoretical models. Must protein folding be a part of a metabolic simulation? It depends on what the simulation is to be used for- since a fraction of metabolism goes towards protein synthesis, we may indeed require a model that can handle the details of protein folding- if for nothing more than accounting for misfolded proteins and the degredation pathway. This could likely *still* be handled by phenomenological models based on experimental results (i.e. fraction of misfolded proteins, etc), but then we can't claim that the model is based on first-principles Physical Law.

What the model 'is' and what information is contained in the model (and here I speak generally, including 'model organisms') entirely depends on what the purpose of the model is- what the results of a simulation/experiment will be used for.

And when you get down to it, nobody is that interested in a MD simulation of a bacterium swimming around. Biomedical research dollars go towards understanding human health and disease.

 

[epenguin]

You might get an appropriate simulation from simulating the things you've mentioned but it won't be useful. If you want to use a simulation to test how an antibiotic works for instance it's vital that you understand the molecular processes.

Fundamentally biology works on the molecular scale, if you want to get meaningful answers of such complex systems you will need to simulate it on that level.

But I am not denying that! The interaction of an antibiotic with a target protein is just normal physics or chemistry. The consequences should be affecting a limited number of processes downstream. So you don't need an equation for the bacterium that is killed and you could say it can be described by physics.

 

[Andy Resnick]

The interaction of an antibiotic with a target protein is just normal physics or chemistry. The consequences should be affecting a limited number of processes downstream. So you don't need an equation for the bacterium that is killed and you could say it can be described by physics.

That's true to some degree, but a predictive ligand-receptor binding model would be invaluable to have. Think about chemotherapy; we currently have drugs designed to target very specific receptors. However, since we have an imperfect understanding, chemotherapy drugs also affect many non-cancerous cells that happen to divide rapidly: gut epithelium, hair follicles, bone marrow.

Or, if you prefer- consider minoxidil and sildenafil citrate. Both chemicals were developed to treat blood pressure via nitric oxide regulation of the renin-angiotensin system. Nobody knew about the side effects, which resulted in both of these chemicals becoming the most prescribed drugs in the history of civilization.

The utility of a predictive model should be clear.