The current pandemic highlights an acute need to develop fast therapeutics against health threats. Traditional approaches to drug discovery are expensive and slow to react to pandemics. In this talk, I will discuss how to accelerate drug discovery with deep learning, and demonstrate their success in antibiotic discovery and COVID-19 drug combination design. In computational terms, the major challenge of drug discovery is molecular graph generation and multi-objective optimization. While deep learning has been extensively investigated for graph encoding, graph generation is a harder combinatorial task and remains under-explored. To address these challenges, I will present novel deep generative models that leverage the low treewidth prior of molecular graphs and demonstrate their success in molecular optimization.