Guest post from Dr. Greg Bowman, University of California, Berkeley
Markov state models
(MSMs) are a powerful approach for investigating the dynamics of proteins and
other biomolecules. The Folding@home team has helped to pioneer the development
of these methods and continues to make important contributions to their further
development. For example, the Huang lab
at the Hong Kong University of Science and Technology has created an exciting
new method for coarse-graining Markov models (the paper is available here).
One of the major
challenges in this area is that the high-resolution Markov models capable of
making quantitative predictions of experiments often have tens of thousands of
parts. As you can imagine, it is hard to
look at each of these parts and the interactions between them to understand the
model. Therefore, it is valuable to
create a new model with, say, a dozen states that still captures much of the
behavior of the more complex model.
The new method from
the Huang lab makes use of a mathematical principle called a Nystrom expansion
to build more accurate coarse-grained models.
The key advantage of this approach is that one can quickly identify the
most important pieces of a model and prove that the remaining pieces can safely
be ignored (or merged into the more important parts). As a proof of principle, the Huang lab has
shown they can build much better models for a few small proteins than is
possible with previous methods.