Structure optimisation by thermal cycling for the hydrophobic-polar lattice model of protein folding

Abstract

Abstract. The function of a protein depends strongly on its spatial structure. Therefore the transition from an unfolded stage to the functional fold is one of the most important problems in computational molecular biology. Since the corresponding free energy landscapes exhibit huge numbers of local minima, the search for the lowest-energy configurations is very demanding. Because of that, efficient heuristic algorithms are of high value. In the present work, we investigate whether and how the thermal cycling (TC) approach can be applied to the hydrophobic-polar (HP) lattice model of protein folding. Evaluating the efficiency of TC for a set of two-and three-dimensional examples, we compare the performance of this strategy with that of multi-start local search (MSLS) procedures and that of simulated annealing (SA). For this aim, we incorporated several simple but rather efficient modifications into the standard procedures: in particular, a strong improvement was achieved by also allowing energy conserving state modifications. Furthermore, the consideration of ensembles instead of single samples was found to greatly improve the efficiency of TC. In the framework of different benchmarks, for all considered HP sequences, we found TC to be far superior to SA, and to be faster than Wang-Landau sampling.

Application of local rules and cellular automata in representing protein translation and enhancing protein folding approximation

“…Approximation algorithms are efficient algorithms that find approximate solutions to NP-hard optimization problems with provable guarantees about the distance of the returned solution to the optimal one. Many evolutionary algorithms have been proposed to tackle the HP folding problem [29], for example, genetic algorithms in combination with different mechanisms [5,46], ant colony optimization [23,28], differential evolution [29], and thermal cycling [19]. Approximation algorithms, on the other hand, provide a guaranteed ratio of the number of 1-1 contacts to the upper bound.…”

Research on predicting 2D-HP protein folding using reinforcement learning with full state space

“…In this model, each amino acid is treated either hydrophobic (H) or hydrophilic (P) and represented as a point on a two-dimensional lattice structure. The rationale behind the HP model is that the hydrophobicity of amino acids is the main driving force for small globulins to form a natural conformation [4]. The primary structure analysis of protein sequences involves the analysis of amino acid physicochemical properties (such as hydrophilicity and hydrophobicity) and sequence patterns, so 2D-HP protein structure prediction refers to predicting the folding structure based on the primary structural analysis of proteins.…”

Recent advances in phase transitions and critical phenomena

“…Irbäck et al [7] review experiments and simulations for the problem of folding in the presence of molecular crowders, where intelligently designed coarse-grained models are required to see the relevant behavior with present-day computational resources. Suitable techniques for determining the folded ground state are discussed for the HP lattice model of proteins by Günther et al [8], focusing on the thermal cycling approach. Polymercolloid mixtures in confinement are the topic of the review by Usatenko et al [9], which discusses the effects of polymer topology as well as different confining potentials, such as slits, with a focus on the field-theoretic treatment of such problems.…”