Rasmus Wedberg
Testing monte carlo algorithms for protein folding simulations
Master Thesis in Theoretical Physics
Abstract:
Simulating the thermodynamics of protein molecules is a
computationally demanding task. Therefore, it is essential to
have efficient simulation algorithms. Here, a few Monte Carlo
algorithms are tested for simulating the thermodynamics of a
short peptide in terms of an all-atom protein model. The
methods are given a certain amount of CPU time and their
efficiency is measured by the statistical error in estimated
expected energy. Three methods, simulated tempering, reduced
dynamical excluded volume and the multicanonical algorithm are
found to be more efficient than the standard Metropolis
algorithm. The results further indicate that for this system,
simulated tempering is the most efficient method of the ones
that were tested. It reduces the simulation time by a factor 7
compared with the Metropolis method.
LU TP 06-45