A new collaborative research accelerator called AI BioDesign is combining artificial intelligence with experimental biology to explore an entirely uncharted design space: the trillions of possible DNA sequences that could exist but never have in nature. Backed by $95 million in funding from the Fund for Science and Technology (FFST), the initiative brings together the Allen Institute, the University of Washington, and Fred Hutch Cancer Center to build AI models that can propose new biological designs, which scientists will then test at scale in a continuous cycle of learning and refinement.

What Makes This Different From Other AI Biology Projects?

While AI is already transforming drug discovery and genomics research worldwide, AI BioDesign takes a distinctly different approach by focusing on open science and reusable resources. Rather than building proprietary tools for a single company or research group, the collaboration is designed to create models, datasets, assays, and tools that the entire scientific community can access and build upon. This emphasis on openness sets it apart from many commercial AI drug discovery platforms that keep their findings proprietary.

The initiative also differs in its experimental grounding. Instead of relying solely on computational predictions, AI BioDesign uses what researchers call a "design-build-measure-learn" cycle. AI models propose new biological designs, scientists physically construct and test those designs in the lab at scale, and the experimental results feed back into the AI models to make them progressively more accurate.

"For the first time, the speed of AI is beginning to match the experimental power of synthetic biology. That changes the question from 'What has nature already made?' to 'What else is possible and how can we test it?'" said David Baker, Nobel laureate and lead scientific director of AI BioDesign.

David Baker, Nobel Laureate and Lead Scientific Director of AI BioDesign, Director of the UW Medicine Institute for Protein Design

How Will AI BioDesign Actually Work?

• Design Phase: AI models analyze patterns in biological data to propose entirely new protein structures and DNA sequences that could theoretically perform useful functions, such as breaking down plastics or fighting cancer.
• Build Phase: Scientists synthesize these AI-designed molecules in the laboratory, testing multiple designs simultaneously using high-throughput experimental methods.
• Measure Phase: Researchers rigorously test whether the designed molecules actually adopt their intended structures, perform their intended functions, and remain stable under biological conditions.
• Learn Phase: Experimental results are fed back into the AI models, which use this real-world data to refine their predictions and become more accurate with each iteration.

This iterative approach represents a fundamental shift in how biological engineering works. Historically, researchers either studied proteins that already exist in nature or tried to modify them through trial and error. AI BioDesign enables what researchers call "de novo design," meaning building entirely new biological structures from scratch.

"De novo design is a new build. We can create exactly the structural features needed for a particular task. We can build molecular elements that have no close natural counterpart. Many times, designing from scratch is actually simpler than trying to engineer around evolutionary compromises found in existing proteins," explained David Baker.

David Baker, Director of the UW Medicine Institute for Protein Design

What Could AI-Designed Biology Actually Solve?

The potential applications span medicine, environmental remediation, and advanced technology. Researchers envision using AI-designed proteins to develop new drugs for cancer and neurodegenerative diseases, create enzymes capable of breaking down ocean plastics, and even build biological computers that consume far less power than silicon-based systems. The initiative also plans to vastly expand the number of genomic datasets available to researchers, enabling AI to identify biological patterns that could inspire solutions to pressing health challenges.

One concrete example already demonstrates the promise of this approach. Researchers at the University of Washington's Institute for Protein Design recently published work in the journal Science showing how AI-designed proteins can solubilize membrane proteins without using harsh detergents, making some of biology's most challenging molecules easier to study. In another project, they created a new class of fluorescent imaging tags called NovoTags that help scientists locate and track specific proteins inside human cells.

What Are the Remaining Challenges?

Despite these advances, significant hurdles remain before AI-designed molecules can reliably treat human disease or solve real-world problems. For a designed molecule to be considered trustworthy, researchers need evidence across multiple levels: that it adopts the intended structure, performs its intended biological function, remains stable under physiological conditions, and continues working when placed in increasingly complex biological environments like living cells and tissues.

The most compelling proof of success will be when an AI-designed molecule functions robustly in actual cells, tissues, organisms, or real-world applications. Equally important is understanding how and why designed molecules sometimes fail or produce unintended interactions, as this knowledge helps paint a complete picture of how a molecule behaves in complex biological systems.

The collaboration brings together complementary expertise across Seattle's scientific ecosystem. The Allen Institute contributes experience building large-scale, open-science platforms; the University of Washington brings expertise in synthetic biology and genome science through institutions like the Institute for Protein Design and the Brotman Baty Institute for Precision Medicine; and Fred Hutch Cancer Center contributes deep knowledge of cellular systems, genomics, and translational medicine.

"What excites me about AI BioDesign is that it brings together the right people and the right institutions at the right time to advance biological design with AI in the loop. The Allen Institute was built for this kind of work: big science, team science, and open science that creates resources entire fields can use," said Rui Costa, president and CEO of the Allen Institute.

Rui Costa, President and CEO of the Allen Institute

The initiative arrives at a pivotal moment in biological research. Advances in single-cell sequencing, subcellular imaging, and computational power have enabled the development of virtual cells that simulate cellular dynamics and predict biological phenomena. Meanwhile, AI is streamlining drug discovery by accelerating target identification and molecular design. AI BioDesign is positioned to complement these efforts by creating openly shared resources that can accelerate biological design across the entire scientific community, potentially transforming how researchers approach some of humanity's most pressing health and environmental challenges.