AlphaFold solved a 50-year-old biology puzzle by predicting protein structures from amino acid sequences in seconds, but Demis Hassabis, CEO of Google DeepMind, says the real work is just beginning. While the 2024 Nobel Prize in Chemistry recognized AlphaFold's breakthrough, Hassabis emphasizes that knowing what a protein looks like is only the first step toward developing actual medicines. The next frontier involves designing chemical molecules that bind precisely to target proteins without triggering harmful side effects from interactions with the body's 20,000-plus other proteins.

What Problem Did AlphaFold Actually Solve?

For more than 50 years, biologists faced a fundamental challenge: predicting a protein's three-dimensional structure from its one-dimensional amino acid sequence. Proteins perform nearly every function in the human body, but their activity depends entirely on their final folded shape, not their chemical composition. When proteins misfold, they can cause diseases including cancer, Alzheimer's, and Parkinson's.

Historically, determining a protein's structure required expensive, time-consuming experimental methods like X-ray crystallography and cryo-electron microscopy. Researchers often spent years and hundreds of thousands of dollars resolving a single structure. Meanwhile, gene sequencing technology advanced rapidly, creating a massive backlog of protein sequences waiting to be analyzed.

Hassabis recognized early that protein folding was fundamentally a pattern-recognition problem, ideal for artificial intelligence. However, when he first encountered the challenge as an undergraduate at Cambridge, AI lacked the necessary capabilities. After DeepMind was founded, the opportunity finally arrived. AlphaFold could not only predict protein structures but do so at astonishing speed, completing predictions in roughly ten seconds per protein.

How Did AlphaFold Change Scientific Research?

Rather than building a traditional online server where scientists submit sequences one by one, Hassabis made a bold decision: compute all known proteins worldwide and make results freely accessible. DeepMind partnered with the European Bioinformatics Institute to create the AlphaFold Database, which now covers nearly all protein structures known to the scientific community and is continuously updated with new gene sequencing results.

The impact has been transformative. More than 3 million researchers worldwide now use AlphaFold, and a pharmaceutical scientist once told Hassabis that he believes almost every new drug in the future will use AlphaFold during its research and development process.

The benefits extend far beyond traditional drug discovery:

• Plant Science: Botanists studying crops with complex genomes now have access to protein structure data that would have taken years to obtain experimentally, enabling faster development of disease-resistant and drought-tolerant varieties.
• Neglected Diseases: Non-profit research institutions studying malaria, Chagas disease, and leishmaniasis can now skip expensive early-stage research and move directly to drug screening, drastically reducing upfront costs.
• Structural Biology: Within less than a year of AlphaFold's release, researchers successfully reconstructed the complete structure of the nuclear pore complex, one of the largest protein complexes in the human body, solving a puzzle that had eluded scientists for decades.

Why Is Protein Design Harder Than Protein Prediction?

While AlphaFold solved prediction, designing new proteins for therapeutic use remains an open challenge. David Baker, the University of Washington researcher who shared the 2024 Nobel Prize with Hassabis, draws a careful distinction: "The reality is that we can now design proteins on a computer. The hype is that for therapeutics, there's a lot more than the basic activity of a protein binding or catalyzing a reaction".

"The reality is that we can now design proteins on a computer. The hype is that for therapeutics, there's a lot more than the basic activity of a protein binding or catalyzing a reaction," said David Baker.

David Baker, Director, University of Washington Institute for Protein Design

Baker's lab published a landmark paper in Nature in November 2025 demonstrating that full-length de novo antibodies can be designed computationally to bind user-specified targets, including the flexible antibody loops that have historically defeated computational design efforts. This represents a genuine capability inflection for antibody drug discovery, which represents a market worth hundreds of billions of dollars.

However, Baker himself identifies the remaining hurdles: manufacturability, in-vivo stability, and immunogenicity remain unsolved at scale. Designing a protein that binds its target is one thing; ensuring it can be manufactured at scale, remains stable inside the human body, and doesn't trigger an immune response is another entirely.

How Are Companies Commercializing Protein Design?

Xaira Therapeutics, launched in 2024 with over 1 billion dollars in total funding, represents the clearest signal that institutional capital believes de novo protein design is fundable at scale. The company was co-founded by Nathaniel Bennett, a former postdoc in Baker's lab, and is led by Marc Tessier-Lavigne, the former president of Stanford and former Chief Scientific Officer of Genentech. The board includes Nobel laureate Carolyn Bertozzi, former FDA Commissioner Scott Gottlieb, and former Johnson and Johnson CEO Alex Gorsky.

This unusually high-density governance structure suggests Xaira is positioning for clinical-stage execution rather than a pure platform licensing play. Baker serves as a scientific advisor to the company.

Beyond therapeutics, Baker's lab is pursuing applications with shorter paths to commercialization, including metallohydrolases for pollutant degradation, programmable nanopores for molecular sensing and sequencing, and biosensor platforms. These areas have lower regulatory and in-vivo stability requirements than therapeutics, making them nearer-term opportunities for de novo protein design.

What's Next for AI-Powered Drug Discovery?

Hassabis acknowledged that AlphaFold is only the first step in drug development. To address the challenge of designing molecules that bind target proteins without off-target interactions, Google's parent company Alphabet founded Isomorphic Labs, which aims to advance from protein prediction to drug design. The new AI system can not only predict protein structures but also design candidate compounds and simulate their binding effects.

For biotech founders and investors evaluating protein design platforms, the concrete technical milestone to anchor due diligence on is Baker's November Nature paper on antibody loop design. Xaira's 1 billion-dollar-plus launch signals institutional confidence in the approach, but therapeutic validation work has not yet been completed publicly.

The broader context is significant: the 2024 Nobel Prize in Chemistry recognized protein structure prediction, while the 2024 Nobel Prize in Physics honored foundational machine learning work by Geoffrey Hinton and John Hopfield. This simultaneous recognition reflects a structural shift in the life sciences. AI is now a first-class methodology, not an accessory tool. For synthetic biology specifically, this accelerates the legitimacy of computational-first design pipelines, where proteins are designed before synthesis rather than discovered through experimental screening.

" }