ptc24: (Default)
Add MemoryShare This Entry
posted by [personal profile] ptc24 at 11:45pm on 09/10/2013
There was a Nobel Prize in Chemistry "for the development of multiscale models for complex chemical systems". This is tangentially related to my PhD, so it seems like something I can have a go at explaining aspects of it.

Proteins often bind to small molecules. Some are receptors that bind to signal molecules, some are enzymes that bind to the transition states of the reactions they catalyse, some are transport proteins, some may have other reasons for binding to things. From my point of view as a sometime supramolecular chemist, there are two questions. Question 1: How come proteins are so big? Question 2: How come the binding constants are so high?

With Question 1 - if you think about all of the bits of the protein that engage in intermolecular interactions with the "guest" molecule, then that's quite small, and if you think about the smallest thing that would tie those bits into a single molecule, that's quite small too. You'd end up with something with a radius... roughly a third of the size of the radius of the average protein maybe? Now maybe proteins are just really inefficient, but that seems unlikely, especially with Question 2. It turns out that we can make artificial "host" molecules for these "guests", often using standard synthetic chemistry techniques, which are nice and space-efficient in that they contain all of the groups that bind to the guest and a small amount of other stuff to make all of those groups into a single molecule (often a big ring with a hole in the middle - a "macrocycle"). There's a problem, though; these synthetic host molecules have much lower host-guest binding constants than their protein counterparts - the binding constants are orders of magnitude lower. So maybe all of those bits of protein are doing something after all.

There was one afternoon in the Chemistry department, when I went to see a research lecture, and they were talking about "QM-MM" - a computational chemistry technique designed specifically to address the issues with proteins binding to small molecules and catalysing reactions. The problem is as follows - you want a really good model of the binding site. To get this, you need to do quantum mechanical (QM) calculations directly. Now you can't solve the equations analytically for anything more complex than a hydrogen atom, and you can only really get numerical solutions that converge exactly for things about as complex as a helium atom (maybe things have moved on a bit since my lectures), but there are some simplifications which make calculations feasible for much larger systems. However, these calculations are still computationally expensive and don't scale well - if you double the number of atoms (really, the number of electrons) in a simulation you more than double the computational demands. So simulating a whole protein like this is a non-starter.

Enter Molecular Mechanics (MM). Here, you back off to classical mechanics. You treat each atom like a ball attached to other atoms via springs; there are some simple equations to say what the relevant forces are for each atom-atom distance. You estimate the parameters for these equations "empirically" - usually this involves doing some QM simulations of small systems, and then fitting the parameters to that. Once you have your MM model set up, you can do calculations *much* faster than QM calculations, and I think they scale better too. One thing you can do with this is Molecular Dynamics (MD) - you calculate the forces on the various atoms at a given time, use that to update the velocities of the atoms, use that to calculate where the atoms will be a few nanoseconds (or whatever) later, then rinse and repeat. You can have a go at simulating protein folding using these techniques, although the last time I was listening to research lectures, it seems there was a long way to go before you could just crank out highly accurate protein structure predictions. One problem with MD is that MD simulations tended to consistently gain or lose energy due to the various approximations involved - not something you want!

So, to try to understand protein hosts binding to small molecule guests (and in some cases catalysing reactions), you'd like to do a simulation. Nothing less than QM calculations will do for the binding site. For the rest of the protein, hopefully you can get away with MM calculations. But what about the bits where the bits you're simulating with QM meet the bits you're simulating with MM. The answer seems to be "lots of complicated maths and physics", and it seems you can get a Nobel Prize for coming up with a sensible way to do this.

So, do these techniques give good predictions to do with proteins binding to small molecules? Insight? Do they help us to explain things? Alas, I don't know the answers to these questions, but we can hope. And, y'know, it's only the Nobel Prize. It's not like having an element named after you - now there's a real honour.
There is 1 comment on this entry. (Reply.)
liv: ribbon diagram of a p53 monomer (p53)
posted by [personal profile] liv at 11:09am on 10/10/2013
Thanks for this. I actually found your explanation more informative than the official lay summary on the Nobel website. All that bizarre stuff with cartoons of Newton crossed with Schroedinger didn't make any sense to me! That may because I'm not a proper lay person, I work in a related field and I do know something about molecular modelling and protein chemistry.


14 15
26 27