Identification of functions of protein sequences requires employment of sophisticated classification algorithms. Support Vector Machines(SVM), rigorously based on statistical learning theory, is once such classification algorithm exhibiting superior
performance. The classifier performance depends upon the ability to provide the most informative set of input features. Classification of data having high dimensionality is usually performed in conjunction with an appropriate feature selection method. We propose an Ant Colony Optimization (ACO)/SVM based hybrid filter-wrapper search technique, for simultaneously extraction of informative features and classification. We evaluate the performance of our algorithm with some important protein function identification problems.

Key words: Protein Functions, Ant colony, SVM-Hybrid.