PARTITIONING THE MOTION IN MOLECULAR-DYNAMICS SIMULATIONS INTO CHARACTERISTIC MODES OF MOTION

Molecular dynamics (MD) simulations result in a comprehensive description of molecular motion. However, to gain insight into the dynamic behavior of molecules it is important to be able to identify different types of motions and characterize them. We have developed a novel technique aimed at characterizing the motion of the system using digital signal processing techniques. The amplitudes and phases of the Fourier transform of the atomic fluctuations are used to define the characteristic modes of motion in the MD trajectory. This yields a pictorial description of the oscillatory motions in a manor analogous to normal-mode (NM) analysis. The validity of this method has been tested on small molecules such as water, acetamide, and a blocked polyalanine in a helical conformation. The NMs obtained by diagonalizing the mass-weighted second derivative matrix were combined to generate ''NM trajectories'' that served as well-characterized test cases. Distinct characteristic modes can be extracted from both NM and MD trajectories. The modes extracted from the NM trajectories were identical to the original NMs. The modes extracted from the MD trajectories were in most cases highly correlated to the corresponding NM. However, intermixing of some of the modes occurred, particularly when conformational changes took place. This technique is flexible and can be applied to the molecular system as a whole or to a subset of atoms of interest. Fourier transform calculations are fast and therefore the analysis stage is not demanding in computational resources. Anharmonicity is included explicitly in the simulations and solvent can be included as well. (C) 1993 by John Wiley & Sons, Inc.