Over the past decade, cryo-EM has experienced a revolution. Triggered by transformative hardware and software improvements, cryo-EM is now able to provide atomic resolution structures for protein complexes routinely. As a consequence, cryo-EM is attracting the interest of the pharmaceutical industry to enable structure-based drug discovery on difficult targets that cannot be studied using X-ray crystallography. However, some essential bottlenecks need to be tackled before the tremendous potential of cryo-EM can be translated into drug discovery. In this project, we aim to address two of these limitations, namely the limited throughput and the occasionally poor quality of ligand reconstructions. By combining classical techniques with modern machine learning methodologies and facilitated by the ever-growing pool of Astex’s cryo-EM protein-ligand structures, this project will deliver a set of new algorithms which will help to drive large-scale drug discovery projects by cryo-EM.