Design and use of the biomolecular interaction network database (BIND) for storing and analyzing protein-protein interaction data

As genomics and proteomics technologies such as mass spectrometry, yeast two-hybrid, phage display and genetic interaction screens become more sensitive and robust, they are becoming more automated and high-throughput. These experimental systems are currently providing a wealth of data on genetic and molecular interactions and post-translational protein modifications. The Biomolecular Interaction Network Database (BIND - http://bind.ca) has been designed to store details about these molecular and genetic interactions, complexes and pathways and thus captures proteomics data in a computer readable format. Chemical reactions, photochemical activation and conformational changes can be described down to the atomic level of detail. Everything from small graph theory methods may be applied for data mining. The database can be used to study networks of interactions, to map pathways across taxonomic branches and to generate information for full pathway kinetic simulations. Currently, BIND is a web-based system that allows the database to be queried and for records to be entered. A Java applet to visually navigate the database and a BLAST against BIND service are both available via the web. BIND is an open community effort. All BIND records are in the public domain and source code for the project is made freely available under the GNU Public License. The system is designed so that both users and a curation staff can submit interactions described in the literature, which are then vetted. BIND has been used to manage and automatically discover new knowledge residing in large yeast protein-protein and genetic interaction networks in Saccharomyces cerevisiae determined using mass-spectrometry, phase-display, yeast two-hybrid and roboticized synthetic lethal screens. A system, called MCODE (Molecular Complex Detection), for automatically recognizing molecular complexes in large molecular interaction networks, has been devised. MCODE is based on the notion that densely connected regions of a molecular network, or graph, represent molecular complexes. The BIND project illustrates how a structured software development process focusing on the design phase provides a sturdy foundation for the future implementation of bioinformatics tools that solve real biological problems.