Unlocking Generative AI for Drug Discovery with Zero-shot Models

Generative AI has the potential to greatly incrase the speed, quality, and controllability of drug design. Traditional discovery requires time and resource intensive screening of large libraries, which also offer little control on the output molecules, which can result in suboptimal candidates. Here, we show how generative deep learning models can produce molecules and representations that could be used for downstream design of new molecules with specified functions.