Predição de estruturas de proteínas utilizando restrições de ângulo diedrais

• Ano de defesa: 2014
• Autor(a) principal: Santos, Karina Baptista dos
• Orientador(a): Dardenne, Laurent Emmanuel
• Banca de defesa: Nicolás, Marisa Fabiana, Souza, Osmar Norberto
• Tipo de documento: Dissertação
• Tipo de acesso: Acesso aberto
• Idioma: por
• Instituição de defesa: Laboratório Nacional de Computação Científica
• Programa de Pós-Graduação: Programa de Pós-Graduação em Modelagem Computacional
• Departamento: Serviço de Análise e Apoio a Formação de Recursos Humanos
• País: Brasil
• Palavras-chave em Português: Biologia molecular; Processamento eletrônico de dados; Algoritmo genético; Molecular biology; Electronic data processing
• Área do conhecimento CNPq: CNPQ::CIENCIAS BIOLOGICAS::BIOQUIMICA::BIOLOGIA MOLECULAR
• Link de acesso: https://tede.lncc.br/handle/tede/204

Protein structure prediction is one of the most important problems in computational molecular biology and aims to determine the three dimensional structure of proteins solely from the amino acid sequence. It is considered that the amino acid sequence contains all the necessary information for the chain to adopt its native conformation under physiological conditions. The prediction of protein structure constitutes an extremely difficult optimization problem because, depending on the approximations made, it deals with hundreds or thousands of degrees of freedom and a complex search space associated with the multimodal energy hyper surface. Even for smaller proteins, the problem is computationally costly due the complexity of the optimization process. The goal of this work was the development of genetic operators for the GAPF program (Genetic Algorithm for Protein Folding – developed by GMSSB/LNCC). The operators work through the use of constraints in the search space of the main chain dihedral angles to accelerate the search algorithm and improve the predictive ability of the program. This work was developed in six stages: I - Building an matrix of dihedral angles phi and psi from a database of experimentally determined protein structures, with up to 20% identity; II - Development and introduction in the GAPF program of the RAMA mutation operators in order to use the dihedral angles matrix; III - Generation of initial population using information supplied by the matrix of dihedral angles; IV - Use of matrix of dihedral angles generated from fragments libraries specifically built for each sequence target; V - Verify the performance of the methodology when information from secondary structure prediction of the target sequence and protein structures database are provided; and VI - Evaluate the effect of these changes in terms of reducing the computational cost and quality of the predicted structures. The operators were tested in a set of eight proteins belonging to the following classes: preferably-alpha, alpha+beta, alpha\beta and preferably-beta. The results showed that the imposition the more favorable phi and psi for each amino acid as angular constraints reduced up to 75% the number of energy function evaluations necessary to obtain equivalent results without the use of these operators. Additionally, it was possible to obtain more accurate models, with a reduction of up to 4Å for RMSD and an increase of approximately 6% in GDT-TS of the bigger proteins of the test set, i.e. 1BDD and 1GYZ, when comparing with the values obtained for standard version of the GAPF program using the same number of evaluations.