Protein Functional Class Prediction Based on Clustering

Protein functional prediction is becoming a challenge in the post-genomic era. Therefore, it is in high demand to develop an automated method which can predict the protein functional class rapidly and accurately. In this paper, we propose a ProfileAA coding and an ExtendProfile coding, then evaluate and compare the two coding methods. Furthermore we choose the ProfileAA coding, which integrates amino acid composition information with amino acid physical and chemical properties information. Moreover, by comparing the coding with three other coding methods, we find that this coding is more reasonable. Next, we predict the protein functional class based on Shortest Path Clustering, combining with nearest neighbor algorithm (NNA). The experimental result shows that our method is more efficient to predict protein functional class.