Algorithmic and Artificial Intelligence Methods for Protein Bioinformatics

Chapter 6

Protein Local Structure Prediction

WEI ZHONG, JIEYUE HE, ROBERT W. HARRISON, PHANG C. TAI, and YI PAN

6.1 Introduction

Studying the protein sequence–structure relationship is one of the most active bioinformatics research areas. A better understanding of the protein sequence–structure correlation can improve effectiveness and efficiency of local protein structure prediction [1]. Many biochemical tests indicate that a protein's sequence can determine that protein's structure completely because all the information that is necessary to specify protein interactions with other molecules is embedded in the protein's sequence [2]. These studies provide experimental support for exploring the protein sequence–structure relationship using data mining techniques. The structure-cluster-based approach, the sequence-cluster-based approach, and the clustering support vector machine (SVM) are used to explore the sequence–structure relationship for local protein structure prediction.

6.2 Structural Cluster Approach

For the structural–cluster approach, protein structural segments are grouped into different structural clusters using multiple structural alignments [3] and unsupervised clustering algorithms [4–6]. Each cluster is associated with a representative local structural prototype. Yang and Wang utilized multiple structural alignments to produce a large set of structure-based sequence profiles [3]. In this multiple structural alignment, sequence segments that have structure similar ...