DFS Based Partial Pathways in GA for Protein Structure Prediction

Abstract

Abstract. Nondeterministic conformational search techniques, such as Genetic Algorithms (GAs) are promising for solving protein structure prediction (PSP) problem. The crossover operator of a GA can underpin the formation of potential conformations by exchanging and sharing potential sub-conformations, which is promising for solving PSP. However, the usual nature of an optimum PSP conformation being compact can produce many invalid conformations (by having non-self-avoiding-walk) using crossover. While a crossover-based converging conformation suffers from limited pathways, combining it with depthfirst search (DFS) can partially reveal potential pathways. DFS generates random conformations increasingly quickly with increasing length of the protein sequences compared to random-move-only-based conformation generation. Random conformations are frequently applied for maintaining diversity as well as for initialization in many GA variations.

An effective evolutionary algorithm for protein folding on 3D FCC HP model by lattice rotation and generalized move sets

“…An initial population was generated randomly from an n - 1 dimensional space within a fixed range. We apply Depth-first search [ 6 , 8 ] to generate random conformations. Each chromosome in the population needs to be evaluated for its fitness value as defined in equation (1).…”

“…Lau and Dill [ 4 ] proposed a simple Hydrophobic-Polar model (HP model) based on the hydrophobic interaction between amino acids which has greatly reduced the complexity involved in protein structure prediction. The HP model has thus been used by many researchers and has been applied in various lattice algorithms [ 5 ] such as 2D Square [ 4 , 6 - 14 ], 2D Triangular [ 15 - 18 ], 3D Cubic [ 13 , 19 , 20 ], 3D Triangular [ 21 ] and 3D Face-Centered Cubic (FCC) [ 22 - 28 ] lattices. Although algorithms on simplified HP lattice methods did not achieve high resolution, they provided a macrocosm-optimized protein structure.…”

DFS-generated pathways in GA crossover for protein structure prediction

“…Thus, low resolution model based ab initio [4] (meaning 'from the origin') or the de novo becomes essential. In an ab initio approach, the building of a 3D conformation (structure) is essentially based on the properties of amino acids, since protein is a three dimensionally folded molecule composed of amino acids [5] linked together (called the primary structure) in a particular order specified by the DNA sequence of a gene [6]. In this paper, our efforts are to investigate the ab initio protein structure prediction problem.…”