Monday, March 6, 2017

A dirty secret in molecular biophysics

The past few decades has seen impressive achievements in molecular biophysics that are based on two techniques that are now common place.

Using X-ray crystallography to determine the detailed atomic structure of proteins.

Classical molecular dynamics simulations.

However, there is a fact that is not as widely known and acknowledged as it should be. These two complementary techniques have an unhealthy symbiotic relationship.
Protein crystal structures are often "refined" using molecular dynamics simulations.
The "force fields" used in the simulations are often parametrised using known crystal structures!

There are at least two problems with this.

1. Because the methods are not independent of one another one cannot claim that a because in a particular case they give the same result that one has achieved something, particularly "confirmation" of the validity of a result.

2. Classical force fields are classical and do not necessarily give a good description of the finer details of chemical bonding, something that is intrinsically quantum mechanical. The active sites of proteins are "special" by definition. They are finely tuned to perform a very specific biomolecular function (e.g. catalysis of a specific chemical reaction or conversion of light into electrical energy). This is particularly true of hydrogen bonds, where bond length differences of less than a 1/20 of an Angstrom can make a huge difference to a potential energy surface.

I don't want to diminish or put down the great achievements of these two techniques. We just need to be honest and transparent about their limitations and biases.

I welcome comments.


  1. I understand that one issue with the classical force fields from a few years ago was their inability to even reproduce accurately such key numbers as the melting/boiling points of ice/water, the anomalous density change etc. Is this still an issue or have the parametrizations improved over time?

    1. There have been improvements. But this is still an issue. It is hard for a single force field to get the whole phase diagram of water correct quantitatively.

      Hence, you should worry even more about how well it is going to describe the interaction of water with a protein.