Hybrid Artificial Bee Colony Algorithm With Enhanced Initialization For Protein Tertiary Structure Prediction
Loading...
Date
2017-08
Authors
Mahmood Alqattan, Zakaria Noor Aldeen
Journal Title
Journal ISSN
Volume Title
Publisher
Universiti Sains Malaysia
Abstract
Protein structure prediction (PSP) is one of the fundamental subjects in bioinformatics applications. The PSP is an NP-complete problem, and is based on predicting the protein native-like structure solely from its primary amino acids sequence. Determining the protein tertiary structure using experimental methods is time consuming and expensive considering that not all protein 3D-structures can be obtained experimentally. Many computational methods have been proposed to overcome the cost and duration of the experimental methods. One of the methods adopted to solve the PSP problem is the ab initio modeling method which has demonstrated great performance results during the last decades. For consideration of optimal prediction results, the ab initio modeling method takes into account three main factors: (1) an accurate energy function, (2) an efficient conformational search algorithm, and (3) the selection of a native-like structure from a pool of conformational structures. This study proposes a new approach for Protein 3D Structure Prediction (PSP) problem through the application of the ab initio method by focusing on the second factor which is to provide an efficient conformational search algorithm, and the third factor which is to use knowledge-based initialization mechanism. This work introduces a modified ABC-GA algorithm using a Crossover operator adopted from Genetic Algorithm, as well as a novel Hybrid ABC with PSO (HPABC) optimization algorithm. These algorithms are used in the conformation searching step of the ab initio method for the PSP problem. The research also
proposes the Angle Range Probability (ARP) list data, which is based on torsion angle classification method and consists of protein torsion angle range probability values. The ARP list is used to enhance the initialization step of the conformational search algorithm in order to improve the search space population and provide more precise conformations. All the proposed methods are evaluated and compared with other well-known applied PSP methods. The experiments are conducted on short sequence benchmark protein called Met-enkephalin that has been intensively used by previous researchers. The proposed ABC-GA and HPABC algorithms achieve significant results in terms of accuracy, time, function evaluation number and success rate. The results also indicate that the best native-like prediction values are achieved from the algorithms that use the ARP list in their initialization step.
Description
Keywords
Hybrid artificial bee colony algorithm , for protein tertiary structure prediction