Computing the Weighted Flow Complex

Joachim Giesen, Matthias John

Proceedings of the 8th International Fall Workshop on Vision, Modeling, and Visualization

Pages 235-243, 2003




Motivation

The flow complex is a data structure that can be used to structure a finite set of points in 1#13 mathend000#. It is closely related to the Delaunay triangulation. It was evident that the flow complex would be a useful data structure for bio-geometric modeling problems. However, the input point sets in bio-geometric modeling problems are weighted points. Thus the theory of flow complexes needs to be extended to include weighted points.

Goal of This Research

The authors wish to extend the notion of a flow complex to that of a weighted flow complex in order to use the flow complex as a data structure in bio-geometric modeling.

Goal of This Paper

The flow complex is basically a data structure that divides up a set of points by defining a distance function based on those points and then classifying regions of space based on which critical point of the distance function each point ``flows" into. All points flow in the direction of steepest ascent (the gradient) of the distance function.

The authors describe how to resolve the complications caused by introducing weighted points in building a weighted flow complex. The basic problem is that each weighted point does not necessarily correspond to a vertex of the flow complex, but each weighted point still can influence the overall structure of the complex. The authors then use the weighted flow complex to decompose biological macromolecules into their components (essentially this means that, given the set of positions of atoms that make up a molecule, predict which pairs of atoms are bonded together) and to give a new definition of pockets in a protein.

Related Work

Results

The authors report the results of computing the flow complex of nine different biological molecules. It appears that the run time per atom is almost constant: for the nine molecules, the average number of milliseconds spent per atom are: 1.47, 2.00, 1.68, 1.74, 1.77, 1.74, 1.77, 1.82, and 1.78.

The authors believe that the definition of pockets given in this paper is more intuitive than the earlier conventional definition.

Bibliography



ctj