Data mining for characterizing protein-protein interaction interfaces

Abstract

The main objective of this thesis is to suggest a new perspective to effectively describe protein-protein interaction interfaces in terms of data mining. Two systematic approaches are applied to answer the question of what features are effective in representing the interaction interfaces. One is proceeded in the aspect of specificity, and the other is performed in the aspect of stability, which are two critical factors determining the binding association. Firstly, specificity of molecular interactions is examined in the context of protein functions. This is the first approach to analyze interaction interfaces at the molecular interaction level in the context of protein functions. We perform systematic analysis at the molecular interaction level using classification and feature subset selection technique prevalent in the field of pattern recognition. To represent the physicochemical properties of protein-protein interfaces, we design 18 molecular interaction types using canonical and non-canonical interactions. Then, we construct input vectors using the frequency of each interaction type in protein-protein interfaces. The 131 interfaces of transient protein-protein heterocomplexes in PDB is analyzed : 33 protease-inhibitors, 52 antibody-antigens, and 46 signaling proteins including 4 cyclin dependent kinase and 26 G-protein. Using kNN classification and feature subset selection techniques, we clearly show that there are specific interaction types according to their functional categories, and such interaction types are conserved through the common binding mechanism, rather than through the sequence or structure conservation. The $C^\alpha-H\cdots O=C$ shows binding specificity for protease-inhibitor complexes, while cation-anion interaction is predominant in signaling complexes. In the case of antibody-antigen complexes, the sign is somewhat ambiguous. From the evolutionary perspective, while protease-inhibitors and signaling proteins have optimized their int...