| Abstract [eng] |
This dissertation describes three novel effective methods for the analysis and evaluation of biomolecular structures. The presented methods construct and utilize the Voronoi tessellation of atomic balls and the tessellation-derived interatomic contact areas. The first method, Voronota, is a method for computing the vertices of the Voronoi diagram of balls. It is capable of processing macromolecular structures efficiently by exploiting common patterns of atomic spatial arrangements. The second method, CAD-score (Contact Area Difference Score), is a highly effective method for the comparison of different conformations of macromolecules, for example, native and modeled structures. It is universally applicable for the comparison of structures of all the major types of macromolecules (proteins, nucleic acids and their complexes). The third method, VoroMQA (Voronoi diagram-based Model Quality Assessment), is a method for the evaluation of predicted protein structures when the native structure is unknown. It efficiently combines the idea of knowledge-based statistical potential with the concept of interatomic contact areas derived from the Voronoi tessellation of atomic balls, it consistently outperforms other statistical potential-based methods. The main conclusion of the presented studies is that Voronoi tessellation-derived contact areas effectively capture important structural features of biological macromolecules. |
