Potent antimalarial drugs with validated activities

Over the past two decades, machine learning methods have excelled at modelling relationships between the structural and physicochemical properties of small molecules and materials and their useful biological, chemical or physical properties. However, until the recent development of generative deep learning methods, it has been almost impossible to reverse the process—that is, use robust and predictive machine learning models to design synthesizable molecules or materials with specific, superior properties. Godinez and colleagues, in Nature Machine Intelligence, recently showed recently how generative methods can be applied to the vital problem of discovering new effective drugs for oft-neglected tropical diseases, specifically malaria.