De novo design of small molecules against drug targets of central nervous system using multi-property optimization

Navneet Bung, Sowmya Ramaswamy Krishnan, Arijit Roy

Abstract: Drug design and development is a time- and cost-intensive process with a low success rate. The process is more complex in case of drugs targeting diseases of the central nervous system (CNS) where blood-brain barrier (BBB) acts as an additional challenge for drug delivery. Recent applications of deep learning in ligand-based drug discovery are promising, although these methods can suffer from lack of target-specific ligand data to train the models. To address these issues, we have developed a de novo drug design method which can design novel molecules by optimizing target specificity as well as multiple properties that make them suitable to cross the BBB. A target-specific ligand dataset is curated by collecting known inhibitors of proteins structurally similar to the target protein. The generative model which learns and designs new molecules is systematically optimized using transfer and reinforcement learning. The reward function is designed to optimize multiple properties simultaneously with state-of-the-art predictive models. The proposed method was validated against the human 5-hydroxy tryptamine receptor 1B (5- HT1B), a G protein-coupled receptor responsible for several psycho-physiological functions and disorders. All existing 5-HT1B inhibitors were collected but used only for validation. We were able to design inhibitors with better binding affinity when compared to the existing inhibitors, with optimized property to cross the BBB. Results from the study show the capability of the proposed method to learn the molecular features required to produce novel small molecules with multiple desired physico-chemical properties against the target protein rapidly.