Dr Aras Asaad presented his work in exploiting existing 2D image-based deep learning architectures to capture 3D molecular information at this year’s AI in Chemistry symposium. 

Established computer vision deep learning architectures require input in a 2D image format. In order to exploit these architectures, Oxford Drug Design has developed a novel representation which captures rich 3D molecular information in a rotationally invariant canonical form as a 2D digital image.