Modern drug discovery has been strongly focused on the development of drugs intended to act against one specific target with high potency and selectivity.. It is now recognized that these concepts may be too simplistic. Designing a single drug molecule able to simultaneously and specifically interact with multiple targets, i.e., show polypharmacological profile is a great challenge of modern drug discovery. At the same time there has been an unprecedented growth of chemical databases incorporating up to several hundred billions of synthetically feasible chemicals. Such Big Chemical Data offer humongous opportunities for discovering novel bioactive molecules. However, the current generation of cheminformatics software tools is not capable of handling, characterizing, and processing such extremely large chemical libraries. Deep learning (DL) method is well suited to address both of these challenges. DL is highly scalable to enable the processing of Big Data, which is in strike contrast to most conventional machine learning approaches. DL also allows for multi-task learning, procedure of learning several tasks at the same time with the aim of mutual benefits. Its architecture allows sharing of learnt information across sub-tasks, and therefore joint training for every endpoint. In this study, we show a construction of structure–activity relationship (SAR) profiles of molecules interacting with multiple biological targets. This is especially challenging when these targets are only distantly related or unrelated, i.e., when they belong to different protein families. We have applied our approach to highly curated chemogenomics data for all major GPCRs and kinase targets annotated in public repositories such as PubChem and ChEMBL. Deployed in our distributed web platform Chembench, these models provide convenient tools for rapid evaluation of off-target bindings, selectivity of novel drug-target associations as well as the identification of novel selective compounds for orphan targets.

Computational drug discovery using deep learning approaches

251st National Meeting (2016)

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