Protein structure prediction in an atomic model with differential evolution integrated with the crowding niching method
- PDF / 3,071,069 Bytes
- 15 Pages / 595.276 x 790.866 pts Page_size
- 75 Downloads / 186 Views
(0123456789().,-volV)(0123456789(). ,- volV)
Protein structure prediction in an atomic model with differential evolution integrated with the crowding niching method Daniel Varela1
•
Jose´ Santos1
Accepted: 8 August 2020 Ó Springer Nature B.V. 2020
Abstract A hybrid version between differential evolution and the fragment replacement technique was defined for protein structure prediction. The coarse-grained atomic model of the Rosetta system was used for protein representation. The high-dimensional and multimodal nature of protein energy landscapes requires an efficient search for obtaining the native structures with minimum energy. However, the energy model of Rosetta presents an additional difficulty, since the best energy area in the landscape does not necessarily correspond to the closest conformations to the native structure. A strategy is to obtain a diverse set of protein conformations that correspond to different minima in the landscape. The incorporation of the crowding niching method into the hybrid evolutionary algorithm allows addressing the problem of the energy landscape deceptiveness, allowing to obtain a set of optimized and diverse protein folds. Keywords Protein structure prediction Differential evolution Evolutionary computing niching methods Crowding niching method
1 Introduction Since the protein native structure is related to its biological function, the prediction of the native structure can be used in the design of specific proteins that interact with other biological molecules, for example, in drug design. Therefore, protein structure prediction (PSP) remains as one of the major challenges in computational biology. There are PSP methods that rely on the knowledge of resolved structures (3D structure is known), stored in databases like Protein Data Bank (PDB) (Protein Data Bank 2003). For example, when predicting the structure of a protein sequence, a common method is the search of proteins with homologous primary structure (amino acid sequence) in the databases, assuming that similar sequences have similar 3D
& Daniel Varela [email protected] Jose´ Santos [email protected] 1
Department of Computer Science and Information Technology, CITIC (Centre for Information and Communications Technology Research), University of A Corun˜a, A Corun˜a, Spain
structure. However, there is an increasing gap between the number (millions) of known protein sequences (result of multiple genome sequencing projects), and the limited number of proteins with resolved structure (hundreds of thousands). This is because wet-lab methods for resolving the location of protein atoms, like X-ray crystallography and nuclear magnetic resonance (NMR) spectroscopy, are expensive, laborious and slow. The ‘‘sequence/structure gap’’ is the reason of the importance of computational ‘‘ab initio’’ methods, that use only the protein primary sequence of amino acids to directly determine the final native structure. In fact, the final structure is assumed to be the one that minimizes the Gibbs free energy and it is also
Data Loading...
