Which minimal order of magnitude is required to simulate single cells and microorganisms in cyberspace?

In an utopian future, people will be possibly able to rapidly design new cures, be their realization a new chemical substance, a nanobot or a synthetic antibody.

"Today we will learn how people built the first super computer which helped to eliminate all human deceases caused by microorganisms", teacher said. "All started with that question on the world building Q&A site.."

Could be this a quote from a hard sci-fi story, if the teacher would go deeper into how that legendary supercomputer had been engineered?

Possibly, a "hard" sci-fi requirement for that would be a need for physically possible supercomputers which can perfectly model and simulate pathogens, more than just a fictional positron brain with endless flops.

Coming to computational performance, as of 2019 we have supercomputers reaching orders of possible operations on peta- and coming close to exascale. Some research is also ongoing to assess what will be required to build a zettascale computer.

But how much is actually enough to have a simulation where a bacteria attacks a human cell and starts to self-replicate?

Relevant figures are:

  • Simulation of viruses like HIV either poliovirus is not a big problem as they have less than 10-100M atoms which seems feasible since some years (which very likely has helped to design the AIDS cure!). Though, a virus is not "life" and there are also "giant" viruses;
  • a bacterial organella has been recetnly simulated leading to a scientific discovery with 100M atoms;
  • an E.coli bacteria has about 1011 atoms;
  • a recent non-biological simulation with fixed molecules allows for simulation of 20x1012 atoms;
  • a "typical" human cell is referred to have 1014 atoms.

The challenges which I could identify so far, but also the according state of the art approaches as of late 2019, look like the following (while atom by atom simulation seems to have good development, anyway):

  • detail of simulation: is Newtonian molecular dynamics simulation enough or is it better to have a quantum-mechanical field interaction which is computationally more expensive?

  • size of simulation: pathogens can differ by orders of magnitude in their size and complexity. This means, between being able to simulate two different species decades might pass. As of today, we can simulate some virus capsids either bacterial organellas on petascale computers.

  • time frame: current petascale simulations allow up to ~45ns/day time frames which is really not much given that many bacteria need hours for one cycle.

That is, will it be anyhow possible, from physical/economical points of view, to build a computer smaller than the Earth and run such simulations?

What I've not considered yet are quantum computers, but today we as it seems actually do not know yet if we will be able to build such systems scaled enough to really compete with large non-quantum ones.

  • 3
    $\begingroup$ A well researched and very interesting question. Could you slip in some worldbuilding context? It would be such a shame to see it closed as off topic. Or possibly consider one of the computer sites, maybe Computational Science could work if it strikes out here. $\endgroup$ Dec 4, 2019 at 12:57
  • $\begingroup$ @WeareMonica thanks! - scicomp.stackexchange.com/questions/33941/… $\endgroup$
    – J. Doe
    Dec 4, 2019 at 13:10
  • $\begingroup$ I'm pretty sure the cure to bacteria will come from bacteriophages filtered out of sewage. The problem is that big medical companies won't research them because they can't be patented. en.wikipedia.org/wiki/Bacteriophage $\endgroup$
    – Fels
    Dec 4, 2019 at 15:00
  • $\begingroup$ @Fels great point! Though I'd like to be sure another one bacteriophag wouldn't take on human cells after finishing off the pathogens. $\endgroup$
    – J. Doe
    Dec 4, 2019 at 15:03
  • 2
    $\begingroup$ @J.Doe you're made of eukaryotes. Bacteriophages work on prokayotes. There's no practical way for them to be able to affect you, too. It is less likely than you falling afoul of a disease that affects potatoes. (I'm assuming you're not a potato-based intelligence, of course) $\endgroup$ Dec 4, 2019 at 17:35

3 Answers 3


It all depends on the accuracy and resolution of your simulation. Ultimately it’s not possible to create a perfect model of a pathogen due to the uncertainty principle operating at the lowest level of resolution in any model.

That said an enormous amount can be learnt from simulations at all scales depending on what you are interested in. The secret is in understanding how to make good approximations for the scale of interest. This is true regardless of the scale because there will always be a level below which you are modelling that has to be approximated. Accurately modelling even a single molecule require approximations of their wave functions and can take enormous amounts of computational power.

  • $\begingroup$ indeed DFT seems to be quite "hot" and of course introduces new scaling problems. "Enormous amounts" how much is that beyond zettascale? how much is enough to see bacteria starting self-replicating in the model like in the real world? $\endgroup$
    – J. Doe
    Dec 4, 2019 at 13:50
  • 1
    $\begingroup$ I think it depends on the resolution you want. What would be the smallest unit in the model? Whatever the level, lower levels must be abstracted. It could be based on large individual molecules, on individual functional groups, on individual atoms or even subatomic particles. Very hard to give you anything definitive but you might find these of some interest: tel.archives-ouvertes.fr/tel-01249604/file/… pdfs.semanticscholar.org/ab0f/… (section 5.2) $\endgroup$
    – Slarty
    Dec 4, 2019 at 17:02

You don't need to simulate the whole organism.

Most bacteria don't "attack" as such, they just...are. That existence produces either waste products that are harmful (think botulism) or trigger the host into some response that is harmful. (think flu causing fever)

So the problem with simulating a disease is not simulating a cell but figuring out what it is interacting with.

So my prediction: we already have the computing power. Big strides will be in automated testing: if bacteria produces enzyme 1, test against proteins A1 through ZZ999, repeat ad infinitum.

This will be done through a similar method that is currently used for building current-state AI, except for "recognize this stop sign" it would be "simulate this protein interaction"

Combine this with automated chemical tests to verify the results and identify new testing-pathways we can start producing the amount of data needed for fast and easy identification of diseases and the best remedy thereof.

  • $\begingroup$ certainly a valid point, still how much scale would be enough for a complete and accurate simulation? $\endgroup$
    – J. Doe
    Dec 4, 2019 at 13:52
  • $\begingroup$ Just one protein interacting with one protein. Couple hundred molecules. Easy peasy. Can do so on my laptop. Now the issue is: which protein should interact with which. $\endgroup$
    – Borgh
    Dec 4, 2019 at 13:55
  • $\begingroup$ that's not an answer to the question. If researchers simulated a whole virus, and now simulate organellas, I think it can be allowed to make a thought experiment and try to extrapolate towards a whole-organism simulation. $\endgroup$
    – J. Doe
    Dec 4, 2019 at 13:56
  • 1
    $\begingroup$ Your question could have been "With the invention of the Dreadnought navies have nothing to fear, how large do you think battleships will be in the future?" which, while a neat question, would prove to be not the right question to ask because someone else had just invented the aircraft. $\endgroup$
    – Borgh
    Dec 4, 2019 at 14:03

The general main challenges with fighting bacterial infections are:

  1. Don’t kill the good guys: First of all, you obviously do not want to kill the human host, but since bacteria are a completely different branch of life, there are lots of bacteria-specific angles of attack here. The bigger problem is that only a small fraction of bacteria is actually harmful, while some of the rest is essential to our survival. For example, each of us carries around kilograms of bacteria in the gut forming a complex interacting microbiome that is relevant to digestion. It is considered the main function of the appendix to store this microbiome in case the body decides to flush the gut (diarrhoea). To make things complicated, some pathogenic bacteria exhibit only very little difference to useful ones.

  2. Evolution: Due to their short life cycle and the abilities to transfer genetic information by means other than reproduction (horizontal gene transfer), bacterial populations can quickly adapt to the environment, including whatever you are using to kill them.

  3. It’s a complex system: Bacterial diseases as well as their cure are a complex process that goes beyond the individual bacterium. Many pathogenic bacteria form multicellular structures (biofilms) that may protect them against drugs and similar. Moreover bacteria interact in many ways with with the human metabolism, the human immune system, other microorganisms, and external factors such as the diet of the host.

  4. Identifying the enemy: Find out which of the aforementioned plethora of microorganisms that inhabitate the host shouldn’t be there.

Our predominant current cure against bacteria are antibiotics, which are roughly speaking poisons that attack all bacteria or a larger subgroup thereof. Antibiotics suffers from all of the above points:

  1. An antibiotic treatment will attack beneficial bacteria and for example can easily have negative side effects on your gut microbiome.

  2. Antibiotic resistance arises due to bacteria evolving to survive the antibiotic.

  3. Many, if not all antibiotic treatments only slow down the bacteria and rely on the human immune system catching up and finishing the job in between.

  4. Your antibiotic won’t work if your disease is due to a kind of bacteria that are naturally immune against it or some totally different type of organism (viruses, fungi, …).

Now why am I telling you all this? The perfect simulation of a single bacterium ignores all of the above points:

  1. It does not simulate the effect of a potential treatment on other bacteria.
  2. It does not explore relevant mutations and their survival (population genetics).
  3. It does not account for the interaction of the bacterium with other bacteria and its environment.
  4. Obviously you need to know your bacterium before simulating it.

Therefore I see no reason to expect that such a simulation will single-handedly solve bacterial diseases. If you want to solve this problem with a brute-force simulation, you would probably at least have to include the entire host with all individual microorganisms inhabitating it. Note that this does not mean that computers and simulations cannot help you with this problem. They can still be used to answer such questions such as:

  • What does the protein encoded by a specific sequence of DNA look like and what does it do (protein folding)?

  • What’s the effect of attacking a specific part of the bacterial metabolism – within the bacterium as well as in interaction with its environment? Note that such simulations can happen on a much higher level than the atomic one.

  • What mutations can arise and how do they affect the bacterial population?

In general, I would not consider computational power a considerable limiting factor at the moment. The main problem is rather the lack of experimental data (what organisms, nutrients, etc. are there?). At least to some extent this is inherent: We cannot perfectly record the nature of every cell (host or bacterium) in a human, let alone without killing them. A considerable amount of scientific work tries to obtain a systematic understanding of the bacterial metabolism, bacterial ecosystems, and antibacterial drugs that can help us to work with limited knowledge.

For whatever it’s worth, I am currently working as a researcher in a group that focuses on microbial resistance and ecology.

  • $\begingroup$ @Wrzprmft thank you very much! I just wonder what would be achievable along this path following the approach of the human brain project where they want to consolidate research results into a unified model. $\endgroup$
    – J. Doe
    Dec 5, 2019 at 10:51

You must log in to answer this question.

Not the answer you're looking for? Browse other questions tagged .