19
$\begingroup$

Today we have huge computational power (which is even significantly larger with supercomputers). I know that computational chemistry is used sometimes to predict particle properties. As I read on Wikipedia:

Present algorithms in computational chemistry can routinely calculate the properties of molecules that contain up to about 40 electrons with sufficient accuracy.

If that's so, why bother to try to find chemical interactions and properties experimentally, at least up to 40 electrons? For example, every year new drugs are being discovered. Wouldn't be it easier at least to find new chemical compounds, if not their properties, simply by computer simulation? What are the constraints and where do they come from? (I know that such constraints exist, but I'd like to know why).

$\endgroup$
  • 14
    $\begingroup$ By the time our computational power grew to the present state, all molecules with up to 40 electrons (which, mind you, constitute a tiny minority of all compounds) were already studied experimentally. As for the new drugs of today, they contain much more than 40 or even 400 electrons, so you can't calculate them reliably. Then again, computational chemistry is an increasingly valuable tool, as long as you know its limitations. Also, welcome to Chem.SE. $\endgroup$ – Ivan Neretin Apr 20 '17 at 14:33
  • $\begingroup$ related chemistry.stackexchange.com/questions/4190/… $\endgroup$ – Mithoron Apr 20 '17 at 15:09
  • $\begingroup$ As R.M. describes in the answer, they do do it, to some extent. en.wikipedia.org/wiki/Docking_(molecular) $\endgroup$ – orthocresol Apr 20 '17 at 15:57
  • 1
    $\begingroup$ It might be worth it to take a quick look at the Travelling Salesman Problem. While it's probably pretty easy to study a molecule and assume it's properties, there are a lot of possible molecules with up to 40 electrons -- The time it would take to go through all of them and list out their properties is presumably pretty difficult. It gets even harder above that. I don't believe the problem would be determining the behavior of an individual molecule type, but searching through them all. $\endgroup$ – Sidney Apr 20 '17 at 17:55
  • 1
    $\begingroup$ @Sidney - 40 electrons is pretty small, so we could easily enumerate all molecules with up to 40e. People have already enumerated (and calculated with QM) essentially all molecules up to 9 heavy atoms: dx.doi.org/10.1038/sdata.2014.22 $\endgroup$ – Geoff Hutchison Apr 21 '17 at 17:34
36
$\begingroup$

Forty electrons is tiny. Even if we limit ourselves to just the valence electons, cyclohexane already has 36 electrons. Anything drug-like has way more electrons that 40. For example, viagra has 178 valence electrons, and that's not necessarily a "large" drug. (Compare with vancomycin, for example.)

Even if you're dealing with things inorganic compounds, where the total number of atoms in the formula unit is small, the properties of the material don't come from a single formula unit, but come from the interaction of a large number of atoms. -- That's an example of a more general principle. The important properties for most materials you use (including drugs) don't come the molecule in isolation, but come from the interactions of the molecule with other molecules, either of the same chemical or of different chemicals. To be accurate, all of those interactions need a system with much more than 40 electrons.

The 40 electron limit comes from the implict assumption here that you're talking about quantum mechanical calculations. QM calculations are rather computationally expensive, as you have to account for all the interactions of all the electrons with each other at all positions in their delocalized superposition. There's various tricks (like DFT) which make the calculations for large numbers of electrons easier, but note that "easier" doesn't mean "easy". Even with DFT and other approaches, large systems take a lot of computer time to calculate accurately.

There are other approaches which don't suffer from the same limit as QM does, but they are able to make their gains in efficiency because they make approximations. For example, molecular mechanics approaches are able to simulate systems in the hundreds of thousands of atoms region. But they're able to do so because they don't actually calculate the position of electrons. Instead they treat the system "classically", and experimentally fit interaction potentials which approximate the underlying quantum effects. (For example, they don't exactly calculate the bond stretching potential, but instead approximate it as a harmonic one. That's "close enough" to the true bond stretching potential for the range of bond lengths typically seen in such simulations, but not 100% quantum mechanically accurate.)

There's many groups and companies which do use molecular mechanics and other similar approaches to inform their drug and material development process. The issue is that because the energetic potentials being used are only approximate the results from the simulation are also only approximate. Depending on what you're trying to simulate, the results of the simulation may or may not be accurate. As such, these simulations are treated mostly as a first step, to find potential leads/hypotheses, and then the scientists actually have to go into the lab and test the results to confirm.

$\endgroup$
  • 7
    $\begingroup$ and those are just the drugs, if you count the drug reseptor sites as well, 40 electrons is like a fart in a desert. Not likely to matter. $\endgroup$ – Stian Yttervik Apr 21 '17 at 6:12
  • 2
    $\begingroup$ I think I'd particularly emphasize the intermolecular interactions and environmental interactions. While QM methods are quite good at treating individual molecules, I think beyond the "40 electrons" question, we're really just starting to tackle interactions. $\endgroup$ – Geoff Hutchison Apr 21 '17 at 17:32
8
$\begingroup$

Aside from the computational power needed to simulate larger molecules, there is also a lack of knowledge about the exact mechanisms that some drugs could potentially use. Think for example of experiments in yeast or Escherichia coli cells, which are used to find new biochemical mechanisms that could be exploited for new drugs. Even though we already know a lot about those cells, it would be computationally very demanding to include all known proteins and mechanisms into any kind of simulation. Furthermore, even if we could do such a simulation, there would still be a whole lot of other proteins, genes and mechanisms which we don't really understand yet but which could very well provide new mechanisms that could be used for new drugs. For this reason we would still need (biological) experiments even if we had much greater computational power than we have today.

$\endgroup$
4
$\begingroup$

Attempt exists, lots of drugs are first generated randomly, those most likely to have useful effects given structural category and simulation (simulation of molecular interactions and of known cell molecular pathways and physiological interactions) are tested, in cell lines, animals, then human. All that has costs and risks.

References about cell simulation:

$\endgroup$

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.