De novo protein design represents the most ambitious frontier of protein engineering: creating functional proteins from first principles without starting from any natural template. David Baker's group at the University of Washington has led this field for decades, using the Rosetta computational platform to design proteins with novel folds, symmetries, and functions. Their achievements include designed protein cages for drug delivery, novel enzyme catalysts, and precisely shaped protein nanostructures that self-assemble with atomic-level accuracy.
The therapeutic potential of de novo proteins is being explored by several companies. Neoleukin Therapeutics, spun out of the Baker lab, designs novel cytokine mimetics that activate specific immune signaling pathways without the side effects of natural cytokines. Generate Biomedicines uses generative AI to design de novo protein therapeutics, including novel antibodies and receptor traps with properties unachievable through optimization of natural scaffolds. These designed proteins can be engineered for superior stability, reduced immunogenicity, and precisely tuned binding properties from the outset, advantages that are difficult to retrofit into natural protein frameworks.
Recent advances in AI-driven generative models have dramatically expanded the scope of de novo protein design. Tools like RFdiffusion from the Baker lab can generate novel protein backbones conditioned on functional constraints, while language model-based generators from EvolutionaryScale and others can produce sequences that fold into stable structures with desired properties. The experimental validation rate for computationally designed proteins has improved from single-digit percentages to the majority of designs folding as predicted, making de novo design an increasingly practical approach for creating proteins with bespoke functions for synthetic biology applications.