Contributions to the study of the protein folding problem using bioinspired computation and molecular dynamics

Abstract

The Protein Folding Problem (PFP) is considered one of the most important open cha- llenges in Biology and Bioinformatics. In this thesis, a novel approach for simulating the protein folding pathways is proposed where, instead using the three-dimensional structure of the protein, the folding states are represented by Contact Maps (CM). A two-dimensional Cellular Automata (2D-CA) evolver is used to simulate the fol- ding process, where each configuration represents a folding state and it is obtained according to its predecessor and a transition rule. Since finding transition rules for simulating a dynamic behavior is a very difficult task, it is proposed a distributed Gene-Expression Programming (GEP)-based approach, called pGEP-CA. Specific fit- ness functions, based on similarity and symmetry measures, are proposed. Futhermore, a heterogeneous parallel Ecology-inspired algorithm is proposed. This algorithm, called pECO, is used for reconstructing the structures from the CMs, using the 3D-AB off-lattice model. Moreover, to the best of our knowledge, it is presented the first application of Molecular Dynamics (MD) to the PFP, using the same model of proteins. Experiments were done to evaluate the adequacy of the proposed approaches. Also, a brief analysis of the load balancing of the parallel architectures is presented. Results show that the approaches obtained coherent results, suggesting their adequacy for the problem. The induced transition rules by the pGEP-CA are able to generate 2D-CA that represent CMs correctly. Concerning the pECO approach, results show that the combination of concurrent evolutionary approaches took advantage of both the coevolution effect and the different search strategies. In addition, it can be observed that the MD approach is capable of displaying biological features such as the hydrophobic core formation and the protein breathing motion. Furthermore, it is observed that parallel processing was not only justified but also essential for obtaining results in reasonable processing time. Finally, concluding remarks and several research directions for future works are presented.