Implementação e análise de modelos de solvatação para a predição ab initio de estruturas de proteínas

• Autor(a) principal: Rocha, Gregório Kappaun
• Data de Publicação: 2011
• Tipo de documento: Dissertação
• Idioma: por
• Título da fonte: Biblioteca Digital de Teses e Dissertações do LNCC
• Texto Completo: https://tede.lncc.br/handle/tede/156
• topic: Estrutura molecular; Bioinformática; Bioinformatics; Molecular structure; CNPQ::CIENCIAS BIOLOGICAS::BIOQUIMICA::BIOLOGIA MOLECULAR

The problem of predicting the native structure of proteins from their amino acid sequence is one of the major challenges of computational biology and implies very high computational cost. Several attempts have been made in the search for efficient algorithms and simplified models for protein structure prediction. In this respect, the inclusion and the correct description of the effects of protein solvent interaction are essential for the success of these methods, considering that the solvent plays a key role in the folding process and structural stability of proteins. Despite recent progress, the modelling of protein-solvent interactions in computer simulations remains a challenge. Implicit solvation models use different strategies to reproduce the effects of the solvent without representing their molecules discretely, doing this with a direct estimate of the solvation free energy. This work aimed to implement and carry out a comparative analysis of implicit solvation models described in the literature, and evaluate the impact of such models in the predictive capacity of the GAPF protein structure prediction suite ( Genetic Algorithms for Protein Folding ), developed in our research group GMMSB/LNCC. As some of the solvation models require the value of solvent accessible surface area (SASA) in their calculations, it was also necessary to implement a method that unites accuracy and computational efficiency. The methodology used in the POPS program was implemented for the calculation of the SASA, and the program MSMS is used as a reference for validation. Four solvation models were analyzed: EAS, I-SOLV, EEF1 e GBobc (used as reference). The thermal unfolding of 15 proteins via molecular dynamics (using the GROMACS simulation package) was carried out to evaluate the solvation models before these were placed in the context of a program for protein structure prediction. Seeking to evaluate the impact of each solvation model on GAPF, large scale tests on the ab initio prediction of structures of a set of 24 proteins were performed. The results show that: (i) the use of the POPS methodology is a good alternative for calculating the SASA; (ii) the solvation models I-SOLV, EEF1 and GBobc reflect the behavior of the SASA in their solvation free energies; (iii) with the exception of EAS, all the models were able to discriminate folding from unfolding structures; (iv) no model was able to discriminate close to native structures from structures with similar compression but with high RMSD (folded incorrectly); (v) the I-SOLV and EEF1 were the solvation models that came closest to the reference model GBobc; (vi) the solvation models I-SOLV and EEF1 provided an improvement in RMSD of predicted structures in the program GAPF with respect to experimental structures; (vii) the I-SOLV and EAS have the lowest computational cost among the evaluated solvation models, being faster than GBobc. The solvation models I-SOLV and EEF1 are the best alternative among those studied to model the effects of the solvation in the protein structure prediction.