Scalable tools for the analysis of chemical compounds using graph-based querying

DESCRIPTION (provided by applicant): The generation, manipulation, storage and retrieval of chemical structures and subsequent calculation of various properties, often related to their biological activity, have become extremely important for drug discovery. The resulting field of Cheminformatics has blossomed in recent years and has been a hotbed for the application of data mining and database principles to collections of chemical compounds. The wide adoption of these techniques has led to im- proved methods for representation of chemical structures, similarity-based retrieval of chemical compounds, diversity analysis, and substructure mining. The representation of chemical compounds as graphs captures the essential aspects of chemical structures in a natural way that can be communicated easily. Recent techniques for graph querying and mining have demonstrated great promise for scalability as well as an improved quality of results over traditional representation techniques such as fingerprints. These techniques include novel ways of graph matching, the organization of graphs in a hierarchical index structure, and the mining of a set of graphs to find statistically over-represented motifs. The proposed research will develop computational tools based on these ideas and investigate the feasibility of the techniques on diverse and large data sets. Graph-based techniques for similar compound retrieval, diversity analysis, and substructure mining will be compared to competing techniques based on other representations of chemical structures. Finally, a system that integrates chemical compound databases with biological databases will be developed. The resulting analysis methods are expected to make a significant impact on the complex, time-consuming, and expensive process of drug discovery. Graph-based representation of chemical compounds results in a more accurate realization of the chemical space. The use of recent techniques in graph querying and mining will enable data analysis that can scale to millions of compounds. The developed system will also integrate information on chemical compounds with biological activity and protein interaction networks, thus enabling more efficient drug discovery.