Multiobjective evolutionary algorithm with many tables for purely ab initio protein structure prediction

Abstract

This article focuses on the development of an approach for ab initio protein structure prediction (PSP) without using any earlier knowledge from similar protein structures, as fragment-based statistics or inference of secondary structures. Such an approach is called purely ab initio prediction. The article shows that well-designed multiobjective evolutionary algorithms can predict relevant protein structures in a purely ab initio way. One challenge for purely ab initio PSP is the prediction of structures with β-sheets. To work with such proteins, this research has also developed procedures to efficiently estimate hydrogen bond and solvation contribution energies. Considering van der Waals, electrostatic, hydrogen bond, and solvation contribution energies, the PSP is a problem with four energetic terms to be minimized. Each interaction energy term can be considered an objective of an optimization method. Combinatorial problems with four objectives have been considered too complex for the available multiobjective optimization (MOO) methods. The proposed approach, called "Multiobjective evolutionary algorithms with many tables" (MEAMT), can efficiently deal with four objectives through the combination thereof, performing a more adequate sampling of the objective space. Therefore, this method can better map the promising regions in this space, predicting structures in a purely ab initio way. In other words, MEAMT is an efficient optimization method for MOO, which explores simultaneously the search space as well as the objective space. MEAMT can predict structures with one or two domains with RMSDs comparable to values obtained by recently developed ab initio methods (GAPFCG , I-PAES, and Quark) that use different levels of earlier knowledge.

Soft Computing Techniques for the Protein Folding Problem on High Performance Computing Architectures

“…In a similar way, the work of Day et al [161] proposes a multiobjectivization for the HP model which scores better results in most of the cases than using a single-objective. Another example of this approach is the work of Brasil et al [162,163]. In this work a new MOEA based on tables, called MEAMT, is presented.…”

Um Estudo Comparativo de Estratégias Evolutivas Aplicadas ao Problema de Roteamento Multicast

“…Assim, espera-se que a questão da diversidade investigada no presente trabalho torne-se ainda mais crítica em cenários com vários objetivos. Diante desse desafio, serão aplicados algoritmos many-objectives [6] para este problema, que trabalham com a combinação dos objetivos [3] e [4], possibilitando que o espaço de busca seja melhor explorado.…”

Selection of computational environments for PSP processing on scientific gateways

“…Research on prediction of protein structures has faced obstacles, despite the evolution of technology [25] . The main objective of researchers is to improve the prediction quality, however, they rarely focus on computational efficiency or reduction of processing time [26] .…”