Artificial Intelligence for Multi-Parametric De Novo Design of GPCR Ligands

While computational algorithms have made great advances recently, their application in the drug discovery context is not trivial, among others due to the lack of availability, and insufficient quality, of biological data about downstream effects of molecules. In this research project, the candidate will now, together with the industry partner Heptares, explore how chemical and biological data can effectively be used in the context of DMTA (Design-Make-Test-Analyze) Cycles, including prospective case studies supported by the company. To this end, both efficacy-related and safety biological readout data such as those from transcriptomics, image-based and histopathology data sources will be combined, which will provide an objective function for compound design. This part will be coupled with a predictive algorithm for structure generation, in order to iteratively provide compound design hypotheses with a rationale for a beneficial efficacy/safety profile in a data-driven manner. This will be applied for different proteins/biological systems both of interest for the industry partner and the academic side, and hence novel ways of generating bioactive chemical matter will be explored and evaluated in real-world situations. The outcome of this project will therefore be a validated novel way of drug design, which, by the integration of relevant biological data in the design process, is able to identify bioactive chemical matter of interest beyond that which would be discovered using simpler design hypotheses, such as those purely on the protein level (where frequently then in vivo behaviour of compounds turns out to be problematic, either due to lack of efficacy or toxicity).