Discovering Statistically and Biologically Significant Clusters of 3D Protein-Protein Interfaces

Presentations by NUS Discovering Statistically and Biologically Significant Clusters of 3D Protein-Protein Interfaces Zeyar Aung [email protected] supervisor: Kian Lee TAN The study of the structural properties of proteinprotein interfaces, which are responsible for interactions of proteins, can give us a better overview of protein functions, as compared to studying individual protein structures separately. In this work, we present a new method to encode, cluster and analyze the similar 3D interface patterns among various protein complexes. We represent the proteinprotein interfaces as 2D residue-residue distance matrices, and encode them as multi-dimensional feature vectors. Then, we cluster the interfaces using these feature vectors, and analyze the resultant clusters by various means. Experimental results show that we can discover a number of statistically significant clusters of interfaces. A visual inspection also confirms that the interfaces that fall into the same cluster are visually similar. We can find out the clusters of similar interface patterns in the protein complexes belonging to diverse structural fold types. We can also discover in some clusters the recurring interface patterns associated with biologically important functional motifs. Furthermore, we compare our method with the sequence-only clustering approach, and observe that ours is much better in terms of the statistical significance of the resultant