Running Head: LOCAL PROTEIN SURFACE CLASSIFICATION Characterization and Classification of Local Protein Surfaces Using Self-Organizing Map

Annotating protein structures is an urgent task as increasing number of protein structures of unknown function is being solved. To achieve this goal, it is critical to establish computational methods for characterizing and classifying protein local structures. We analyzed the similarity of local surface patches from 609 representative proteins considering shape and the electrostatic potential, which are represented by the 3D Zernike descriptors. Classification of local patches is done with the emergent self-organizing map (ESOM). We mapped patches at ligand bindingsites to investigate how they distribute and cluster among the ESOM map. We obtained 30-50 clusters of local surfaces of different characteristics, which will be useful for annotating surface of proteins.