Simpler model gets to the point with proteins

Rice University scientists introduce a technique to study dynamic processes with greater accuracy.


Rice University have recently developed Observable-driven Design of Effective Molecular Models (ODEM), that can more accurately regenerate experimental results with simple coarse-grained models used to simulate protein dynamics. The framework process the data in the definition of a coarse-grained simulation model.

The model can reveal unanticipated molecular properties. During experiments, the researchers discovered a new detail about the folding mechanism of FiP35, a common WW domain protein that is a piece of larger signaling and structural proteins.

The Rice lab of computational chemist Cecilia Clementi said, “Understanding proteins, especially their dynamics, is essential to understanding life. There are two complementary ways to do this: either through simulation or experimentation. In an experiment, you measure something that’s real, but you’re very limited in the quantities you can measure directly. It’s like putting together a puzzle with only a very few pieces.”

Rice Professor Cecilia Clementi and graduate student Justin Chen. (Credit: Rice University)
Rice Professor Cecilia Clementi and graduate student Justin Chen. (Credit: Rice University)

“The simulations allow researchers to look at every aspect of protein dynamics, but models that incorporate the properties of every atom can take supercomputers months or years to compute, even if the proteins themselves fold in seconds in vivo. For faster results, scientists often use coarse-grained models, simplified simulations in which a few effective “beads” represent groups of atoms in a protein.”

“In very simple models you have to make strong approximations, and as a consequence, the results may differ from reality. We combine these two approaches and use the power of simulation in a way that reproduces the experiments. That way, we get the best of both worlds.”

Rice graduate student and lead author Justin Chen said, “Acquiring initial data is not an issue. There is a wealth of experimental data about proteins already, so it’s not hard to find. It’s just a matter of finding a way to model that data in a simulation.”

The framework uses data from one or a combination of sources like Förster resonance energy transfer (FRET), mutagenesis or nuclear magnetic resonance. Benefitting by  Markov models, it merges multiple short protein simulations to obtain the equilibrium distribution of protein configurations that is used in ODEM.

Clementi stated, “In our simplified models, we include only the physical factors we think are important. If by using ODEM the simulations improve their agreement with experiments, it means that the hypothesis was correct. If they do not, then we know there are ingredients missing.”

Chen said, “Now we’re scaling it up to larger systems, like 400-residue proteins, about 10 times larger than our test protein. You cannot do full-atom simulations of these large motions and long timescales, but if you do 10 or 11 iterations of a coarse-grained model with ODEM, they take only a few hours. That’s a huge reduction of the time it would take a person to see reasonable results.”

Their work appears in the American Chemical Society’s Journal of Chemical Theory and Computation. Co-authors of the paper are former Rice undergraduate Jiming Chen (now at the University of Illinois Urbana-Champaign pursuing a Ph.D. in chemical engineering) and Giovanni Pinamonti, a postdoctoral researcher at the Free University of Berlin. Clementi is a professor of chemistry and of chemical and biomolecular engineering, Rice’s Wiess Career Development Chair in the Department of Chemistry, a senior scientist at the Center for Theoretical Biological Physics at Rice, the Einstein Visiting Fellow at the Free University in Berlin and co-director of the National Science Foundation (NSF)-supported Molecular Sciences Software Institute.