Synthetic Biology Meets AI to Design New Molecules

How AI-Driven Design Rewrites Biological Engineering

For decades, researchers followed a strict linear path from sequence to structure to function when studying proteins. Today, advanced artificial intelligence allows scientists to reverse that trajectory. According to Nobel laureate Prof. David Baker, lead scientific director of AI BioDesign and director of the UW Medicine Institute for Protein Design, researchers can now start with a desired function—such as binding a target or catalyzing a reaction—and work backward to identify the necessary protein structures and DNA sequences.

Natural evolution produced an astonishing variety of life over billions of years, yet that diversity barely scratches the surface of possible DNA sequences. Modifying natural proteins often resembles renovating an old building, according to Baker, because researchers must work around evolutionary compromises. In contrast, de novo design builds molecular elements completely from scratch, allowing scientists to create exact structural features for specific tasks without natural counterparts.

Continuous Design-Build-Learn Cycles

The AI BioDesign platform operates on continuous design-build-measure-learn cycles supported by the Fund for Science and Technology. AI models propose entirely new biomolecules that researchers then build and test at scale. Experimental results flow directly back into the computational models, sharpening their accuracy over time.

Modern machine learning models learn patterns linking sequence, structure, and function to generate candidate proteins that random screening would likely miss. Baker and his team at the Institute for Protein Design recently demonstrated this potential in studies published in the journal Science. Researchers introduced a method to solubilize membrane proteins without detergents using AI-designed proteins and created NovoTags, a new class of fluorescent imaging tags designed to locate specific proteins and track interactions inside human cells using advanced light microscopy.

Translating Novel Molecules into Real-World Solutions

While single-cell sequencing, subcellular imaging, and virtual cells simulate cellular dynamics and streamline drug discovery across the research landscape, translating AI-designed molecules into practical tools presents distinct hurdles. Before deployment in human health or sustainable manufacturing, designed molecules must prove they adopt the correct structure, perform the intended biological function, and remain stable under physiological conditions.

Biological systems operate as complex networks rather than isolated parts, according to Baker. Predicting how custom molecules behave within these intricate environments remains a primary challenge for the field. Moving forward, bridging computational design, high-throughput experimentation, and real-world deployment will determine how effectively these innovations translate into medicines, diagnostics, and sustainable technologies.

Frequently Asked Questions

What is the AI BioDesign accelerator?

AI BioDesign is a collaborative research accelerator uniting the Allen Institute, the University of Washington, and the Fred Hutch Cancer Center to combine artificial intelligence and synthetic biology for engineering novel biomolecules.

How does de novo protein design differ from traditional genetic engineering?

Traditional engineering modifies existing natural proteins shaped by evolution, whereas de novo design builds completely new molecular structures from scratch to achieve specific desired functions.

What are some potential applications of AI-designed proteins?

Applications include developing drugs for neurodegenerative diseases, creating plastic-destroying enzymes, building more power-efficient biological computers, and designing advanced fluorescent imaging tags like NovoTags.