Multi-resolution protein model

Abstract

The area of molecular biology opens new applications for the communities of computer graphics, geometric modeling and computational geometry. It has been a usual understanding that the structure of a molecule is one of the major factors determining the functions of the molecule and therefore the efforts to better understand the molecular structure have been made. It turns out that the analysis and the prediction of the spatial structure of a molecule usually takes a significant amount of computation even though the number of atoms involved in the molecule is relatively small. Examples are the protein-ligand docking, protein folding, etc. In many molecules, however, the number of atoms is quite large. The number of atoms in the system varies from hundreds to thousands of thousand. The problem size gets even larger by both incorporating more details of a model and expanding the scope of the model from a single protein to a whole cell. This trend will continue as the computational resource gets more powerful and therefore the computational requirement will always remain critical. In this paper, we propose a multi-resolution model for a protein (MRPM) to find a seemingly optimal trade-off between the computational requirement and the solution quality. There are two aspects of the proposal: The avoidance of computation and the delay of computation until it is really necessary.