Data mining of protein sequences with amino acid position-based feature encoding technique

Biological data mining has been emerging as a new area of research by incorporating artificial intelligence and biology techniques for automatic analysis of biological sequence data. The size of the biological data collected under the Human Genome Project is growing exponentially. The available data is comprised of DNA, RNA and protein sequences. Automatic classification of protein sequences into different groups might be utilized to infer the structure, function and evolutionary information of an unknown protein sequence. The accurate classification of protein sequences into family/superfamily based on the primary sequence is a very complex and open problem. In this paper, an amino acid position-based feature encoding technique is proposed to represent a protein sequence using a fixed length numeric feature vector. The classification results indicate that the proposed encoding technique with a decision tree classification algorithm has achieved 85.9% classification accuracy over the Yeast protein sequence dataset.