0.8 Motion planning for proteins: biophysics and applications  (Page 4/10)

For more information on CHARMM and the CHARMM force field, please see The CHARMM website.

Applications of roadmap methods

Kinetics of protein folding

The two standard methods of simulating protein motion are molecular dynamics simulation (MD) and Monte Carlo simulation (MC). In MD, a molecule or system of molecules is given an initial set of atomic momenta, placed in a potential field, and allowed to evolve over time following Newton's equations of motion and the forces exerted on it by the field. In MC, a series of perturbations is applied to a single molecule. After each perturbation, if the estimated energy of the molecule has decreased, the perturbed conformation is used for the next step of the simulation. If the energy has increased, the perturbed conformation might be accepted, with a probability that drops off sharply as the energy change increases. Otherwise, the perturbed conformation is rejected and the previous conformation is perturbed again. Properly implemented MC or MD simulations, run for long enough, should generate a series of conformations with a Boltzmann-like distribution of structures (see the first section of this module for a reminder of what the Boltzmann distribution looks like).

The problem with these methods is that they are very slow. A single MD simulation of a few nanoseconds of motion for an average-size protein, performed on a cluster of processors, can take days, and such simulations are of limited reliability due to approximations of energy and to the extremely short time periods that can be simulated in a reasonable amount of CPU time. Simulations must be repeated to determine what a reasonable, average behavior might look like. Some protein rearrangement events take place on a scale of microseconds, milliseconds, or even seconds, so a trajectory of a few nanoseconds cannot hope to capture these low-frequency events.

The field of chemical kinetics is concerned with the rate at which chemical processes take place, and therefore, the pathways and mechanisms by which they occur. In protein biochemistry, one of the major open questions is the protein folding problem: Given a protein, and its folded (native) and unfolded (denatured) structures, what is the mechanism by which the protein folds into its native state? Currently (2006), it is possible to determine in the laboratory the rate at which a protein folds and sometimes the form of its transition state, the highest energy conformation(s) it assumes in the process of folding. These laboratory measurements can be compared to those inferred from simulation, and the quality of the simulation can thereby be indirectly estimated.

Roadmap-based algorithms to study this problem began with work by Latombe, Singh, and Brutlag in 1999 , in which they attempted to use a PRM to find and study protein binding pockets. This work led directly to that of Song and Amato, Apaydin and Latombe, and Singal and Pande, all presented below. Existing methods as of early 2006 are presented.