Visual Analysis of Atomic Structures Based on the Hard-Sphere Model

• Visualization and Analysis of atomic compositions is essential to understand the structure and functionality of molecules. There exist versatile areas of applications, from fundamental researches in biophysics and materials science to drug development in pharmaceutics. For most applications, the hard-sphere model is the most often used molecular model. Although the model is a quite simple approximation of reality, it enables investigating important physical properties in a purely geometrical manner. Furthermore, large data sets with thousands up to millions of atoms can be visualized and analyzed. In addition to an adequate and efficient visualization of the data, the extraction of important structures plays a major role. For the investigation of biomolecules, such as proteins, especially the analysis of cavities and their dynamics is of high interest. Substrates can bind in cavities, thereby inducing changes in the function of the protein. Another example is the transport of substrates through membrane proteins by the dynamics of the cavities. For both, the visualization as well as the analysis of cavities, the following contributions will be presented in this thesis: 1. The rendering of smooth molecular surfaces for the analysis of cavities is accelerated and visually improved, which allows showing dynamic proteins. On the other hand, techniques are proposed to interactively render large static biological structures and inorganic materials up to atomic resolution for the first time. 2. A Voronoi-based method is presented to extract molecular cavities. The procedure comes with a high geometrical accuracy by a comparatively fast computation time. Additionally, new methods are presented to visualize and highlight the cavities within the molecular structure. In a further step, the techniques are extended for dynamic molecular data to trace cavities over time and visualize topological changes. 3. To further improve the accuracy of the approaches mentioned above, a new molecular surface model is presented that shows the accessibility of a substrate. For the first time, the structure and dynamics of the substrate as hard-sphere model is considered for the accessibility computation. In addition to the definition of the surface, an efficient algorithm for its computation is proposed, which additionally allows extracting cavities. The presented algorithms are demonstrated on different molecular data sets. The data sets are either the result of physical or biological experiments or molecular dynamics simulations.

$Rev: 13581$