In Bioinformatics Protein Fold Recognition (PFR) and Structural Class Prediction (SCP) is a significant problem in predicting protein with a three dimensional structure. Extraction of valuable features of protein that consists of 20 amino acids to acquire more desirable classifiers is fundamental to this PFR and SCP. Feature extraction technique predominantly exploits Forward Consecutive Search Scheme (FCS) that supplements syntacticalbased, evolutionary-based and physicochemical-based information. In this research work, a classifier known as Enhanced Artificial Neural Network (ANN) is employed as it is more efficient than Forward Consecutive Search scheme in order to improve the performance of PFR and SCP. The Enhanced ANN algorithm is an improved version of Artificial Neural Network when compared with various existing algorithms such as Support Vector Machine (SVM), ANN, K-Nearest Neighbor (KNN) and the Bayesian. The experiments are conducted on four datasets namely DD, EDD, TG and RDD. Ultimately, the statistical imputation of Enhanced ANN algorithm hypothesizes gives better results than other algorithms to improve the performance of PFR and SCP.

Introduction

Proteins are the components which play important roles in the activities of organisms. Protein’s function depends on the interactions with other proteins and its folding. Mismatch protein folding usually leads to changing in properties of the protein, which causes some diseases (Hashemi et al., 2009). To acquire knowledge about the protein function, interactions and regulations the prediction of protein structural classes is extremely useful (Jian-Yi Yang et al., 2010). To Increase the prediction accuracy of secondary structure and also to reduce the testimony of hunting scope in three dimensional structure predictions, the mastery of the structural class is helpful (Mohammad and AliYaghoubi, 2016). The SCP has become one of the most important features for characterizing the overall folding type of a protein in protein research. The first definition of protein structural class was introduced by Levitt and Chothia in 1976 and the globular proteins are normally classified into four structural classes such as (i) the all-α class consists of only little amount of strands, (ii) the all-β class consists of only little amount of helices, (iii) the α/β class consists of helices and almost all parallel strands, and α + β class consists of helices and almost all anti-parallel strands (Levitt and Chothia, 1976). Basically, the structural class of protein prediction from 20 amino acids is a significant task in the field of molecular biology. Proteins with unique length and similarities to be a part of the same fold having the identical significant protein secondary structure in the identical arrangement with the identical topology certainly they have a regular origin of evolutionary (Yang et al., 2011). PFR is used to model the proteins which have the similar fold as proteins of known structure, but do not have homologous proteins with known protein structure. PFR is the acquiring of three dimensional structure of the protein sequences independent from the sequence identities (Ding and Dubchak, 2001). PFR and SCP are prohibited as a transitional step for identifying the protein three dimensional structures. The PFR and SCP consist of two main concepts such as feature extraction techniques and classification techniques. The main goal of PFR and SCP is to allocate the novel protein sequence to a particular fold type and to a particular class type. Computational approaches considered more attention over the years due to the expense and the time involved in identifying the three dimensional structure of protein by using X-ray crystallography and Nuclear Magnetic Resonance (NMR) (Ibrahim and Abadeh, 2017).