Automatic Classification of Protein Structures Based on Convex Hull Representation by Integrated Neural Network
The large scale deposited data and existing manual classification scheme make it possible to study the automatic classification of protein structures in machine learning framework. In this paper the classification system is constructed by an integrated feedforward neural network through incorporating the expert judgements and existing classification schemes into the learning procedure. Since different aspects of a protein structure may be relevant to various biological problems, the protein structure is represented by the convex hull of its backbone and geometric features are extracted. The training and prediction tests for different training sets in the class level of CATH indicate that the new automatic classification scheme is effective and efficient. Also the neural network model outperforms hidden markov model and support vector machine in the comparison experiment.
KeywordsSupport Vector Machine Convex Hull Hide Markov Model Training Accuracy Validation Accuracy
Unable to display preview. Download preview PDF.
- Baldi, P., Brunak, S.: Bioinformatics: The Machine Learning Approach. MIT Press, Cambridge (1998)Google Scholar
- Hadley, C., Jones, D.T.: A Systematic Comparison of Protein Structure Classifications: SCOP, CATH and FSSP. Structure with Folding & Design 7, 1099–1112 (1999)Google Scholar
- Zhang, X.S., Zhan, Z.W., Wang, Y., Wu, L.Y.: An Attempt to Explore the Similarity of Two Proteins by Their Surface Shapes. In: Operations Research and Its Applications, Lecture Notes in Operations Research, vol. 5, pp. 276–284. World Publishing Corporation, Beijing (2005)Google Scholar
- Chang, C.C., Lin, C.J.: LIBSVM: A Library for Support Vector Machines (2001), http://www.csie.ntu.edu.tw/cjlin/libsvm
- Young, S.: The HTK Book version 3, Microsoft Corporation (2000)Google Scholar