Exploring sets of molecules from patents and relationships to other active compounds in chemical space networks

Ryo Kunimoto,Jürgen Bajorath

DOI: https://doi.org/10.1007/s10822-017-0061-2

Abstract:Patents from medicinal chemistry represent a rich source of novel compounds and activity data that appear only infrequently in the scientific literature. Moreover, patent information provides a primary focal point for drug discovery. Accordingly, text mining and image extraction approaches have become hot topics in patent analysis and repositories of patent data are being established. In this work, we have generated network representations using alternative similarity measures to systematically compare molecules from patents with other bioactive compounds, visualize similarity relationships, explore the chemical neighbourhood of patent molecules, and identify closely related compounds with different activities. The design of network representations that combine patent molecules and other bioactive compounds and view patent information in the context of current bioactive chemical space aids in the analysis of patents and further extends the use of molecular networks to explore structure-activity relationships.