The number of high-quality protein-ligand structures is increasing every year and opens the route for a detailed knowledge discovery needed for various structure-based molecular design tasks. Especially the analysis and comparison of spatial arrangements of atoms (interaction patterns) in large sets of protein-ligand interfaces can help deepening the understanding of molecular recognition and improving the design of highly affine ligands. The deduction of such information however requires efficient databases mining systems which are capable of dealing with the complexity of varying spatial queries in reasonable time.<br/>To address this need we developed a new database and connected retrieval system which is able to mine large sets of protein-ligand complexes for specific interaction patterns. Highly efficient indexing techniques and different graph-based algorithms are used to rapidly detect all occurrences of a spatial query. The query consists of several atom descriptions which are connected by distance intervals or precompiled interactions. An atom description can be further refined by a molecule type and a substructure it belongs to. Additionally, constraints for physico-chemical properties can be combined with the spatial query. A graphical user interface allows an intuitive definition of queries starting from scratch or from a known binding site. Results can easily be analyzed and compared in a 3D viewer showing the structures aligned to the query.<br/>In this contribution we will present the structure of the spatial queries used in our approach and explain in detail which indexing techniques and algorithms we used to efficiently mine large sets of protein-ligand structures.

Indexing techniques and algorithms to efficiently mine interaction patterns in large sets of protein-ligand-complexes

251st National Meeting (2016)

The official archive and repository for all ACS Meetings & Expositions abstracts and select posters or presentations.