One aim of current research is to find characteristics of protein-protein interactions in order to categorize them. For the analysis of these interactions we have to deal with a large amount of data. Therefore we decided to construct a new database that collects all desired interaction features and allows a comfortable search via web interface.
As protein-protein interactions are one of the basic mechanisms in most cellular processes, it is desirable to understand the molecular details of protein-protein contacts and ultimately be able to predict which proteins interact. Interface areas on a protein surface that are involved in protein interactions exhibit certain characteristics. Therefore, several attempts were made to distinguish protein interactions from each other and to categorize them.
One way of classification are the groups of transient and permanent interactions. Examples of properties are the amino acid and secondary structure element composition and pairing preferences. Certainly, interfaces can be characterized by many more possible attributes and this is a subject of intense ongoing research.
Although several freely available online databases exist that illuminate various aspects of protein-protein interactions, we decided to construct a new database collecting all desired interface features allowing for facile selection of subsets of complexes. As database-server we applied MySQL and the program logic was written in JAVA. Furthermore several class extensions and tools such as JMOL were included to visualize the interfaces and JfreeChart for the representation of diagrams and statistics.
The contact data is automatically generated from standard PDB files by a tcl/tk-script running through the molecular visualization package VMD. Currently the database contains 536 interfaces extracted from 479 PDB files and it can be queried by various types of parameters. The next step is to apply machine learning techniques to classify protein-protein complexes.