De Novo Drug Protein Design for Drug Discovery - Natalie Tatum (Newcastle University, UK)

Recent advances in protein structure prediction have revolutionized how we model individual proteins and their interactions within complexes. Tools such as co-folding now allow us to predict multi-protein complex assemblies with defined confidence. Building on this foundation, diffusion-based generative models extend this to de novo protein design by iteratively refining random, noisy backbones into realistic structures, guided by learned structural relationships and user-defined constraints. This opens the door to designing proteins with tailored functions, including high-affinity binding to a chosen target. Our focus is on the design of protein binders as tools for drug discovery, particularly to probe target biology and validate sites for small-molecule development. In this talk, we will outline the principles behind de novo protein design, explain practically how designs can be scaffolded and guided for site-specific binding, and discuss the methods for evaluating, scoring and selecting designs for experimental validation through examples from the Newcastle Drug Discovery Group exploring cyclin–CDK biology.