Abstract
Protein structure comparison (PSC) and classification have been utilized to comprehend evolutionary relationship between protein structures and functions. However, PSC is computationally time-consuming due to the multiple dimensions of geometric information and the complexity of spatial organizations of atoms. In order to reduce the computational complexity, we have developed a novel PSC method by transforming a three-dimensional structure into a two-dimensional angle-distance (A-D) image. By converting geometric comparison problems into image template matching problems, our methodology not only achieves an improved PSC efficiency but also brings about some unique properties and applications that are difficult for conventional PSC methods. Angle-distance images are created by utilizing secondary structure information of proteins. Subsequently, they are compared by using the cross-correlation approaches which are free from the limitations of the connectivity of secondary structural elements (SSEs) and the spatial orientations of individual domains. Similarities between protein structures are thus identified as various similarities of patterns in A-D images. Our experimental results demonstrate that the proposed method can accurately and efficiently classify protein structures at the fold level defined by the SCOP database even for proteins sharing low sequence identities. Based on the A-D image techniques developed here, we develop a novel and the first practical detection method for three-dimensional domain swapping (DS). DS is a mechanism for forming protein quaternary structure that can be visualized as if monomers had “opened” their “closed” structures and exchanged the opened portion to form intertwined oligomers. DS has been considered possible to occur in a protein with an unconstrained terminus under appropriate conditions. It may play important roles in the molecular evolution and functional regulation of proteins, and in the course of formation of Alzheimer’s and prion diseases. In addition, DS is promising for the design of auto-assembling biomaterials. Given the increasing interest paid to DS and the lack of bioinformatics resources specifically designed for studying DS, our developments may help move related fields forward. To sum up, in this dissertation, a new PSC methodology and a novel detection system for DS have been proposed. The results have been demonstrated to be more applicable to detect DS relationships than the well-known existing sequence/structural alignment and domain motion detection methods. In the future, DS database is expected to be built to promote the development of the related biological engineering.