International Journal of Computational Intelligence Research (IJCIR)

Volume 2, Number 1 (2006)


Exploring the classification of protein structures on geometric patterns 

by neural networks

Wang Yong, Yong Wang, Ling-Yun Wu, Xiang-Sun Zhang, Luonan Chen
Osaka Sangyo University, Nakagaito 3-1-1, Daito, Osaka 574-8530, Japan
Academy of Mathematics and Systems Science, CAS, Beijing 100080, China


The need for automatic classification of protein structures is urgent since different aspects of a protein structure may be relevant to various biological problems and the growth of large-scale protein structure database is rapid. In this paper the problem of structural classification of protein database is explored in the machine learning framework by neural networks to incorporate the expertsí judgments and existing classification schemes into the learning procedure. Firstly, the geometric patterns are extracted from each protein structure. Then the classification system is constructed by a feed-forward neural network. The preliminary training and prediction tests in class level for different feature and training sets confirm the effectiveness and efficiency of the new automatic classification scheme, and imply the possibility to classify protein structures from geometric viewpoint.

protein structure, automatic classification, neural network, geometric pattern.