Early-stage drug discovery has long relied on screening strategies, in which potential small-molecule binders are identified through high-throughput searches of available collections of chemical compounds. While widely used in practice, this approach is inherently limited by the available molecular libraries, hindering its applicability to highly constrained targets such as protein-protein interactions. Moreover, it often yields non-specific and weak binders requiring laborious hit-to-lead optimization campaigns. In my work, I explore an alternative strategy : generative structure-based drug design. Instead of searching for the best small-molecule candidate among existing ones, deep generative models learn the underlying distribution of molecular data and design custom, chemically novel compounds for a specific protein in a matter of seconds. In this talk, I will introduce the key concepts of generative drug design, highlight its current limitations, and discuss its therapeutic potential.​

​