AI creates for the first time an experimental treatment for a disease forgotten by pharmaceutical companies

David Baker, winner of the latest Nobel Prize in Chemistry, claims that humanity is undergoing a transformation as momentous as learning to handle metals at the end of the Stone Age. He speaks of “the protein design revolution,” comparable to the Industrial Revolution, which changed the planet with its steam engines. His laboratory at the University of Washington announced on Wednesday that its disruptive artificial intelligence programs, for which Baker won the Nobel Prize, have succeeded for the first time in creating a treatment for a disease forgotten by big pharmaceutical companies. At the forefront of this scientific feat is the Mexican biochemist Susana Vázquez, who has just left the United States to join the National Cancer Research Center in Madrid, Spain.

Baker’s lab invents proteins that don’t exist in nature. A couple of years ago, its members presented “the world’s first computationally designed protein medicine”: a vaccine against Covid, called SKYCovione, which is already in use in the United Kingdom and South Korea. The researchers have also created some promising molecules against the flu and brain cancer. In her final year of her PhD, Vázquez proposed trying one of the 23 neglected tropical diseases according to the World Health Organization: snakebite envenoming, which causes more than 100,000 deaths a year and three times as many amputations. She and her colleagues used RFdiffusion and ProteinMPNN, two artificial intelligence programs that designed previously nonexistent proteins capable of neutralizing the deadly toxins in cobra bites, at least in computer simulations.

Vázquez, born in Querétaro 31 years ago, was jogging in Seattle when she received an email with the first results of the animal experiments. “My heart stopped and I had to stop to read the email. It was super exciting, because some of the mice had survived 100% of lethal doses of venom,” she recalls. Her study was published this Wednesday in the journal Nature, a showcase of the best science in the world. The authors believe that, beyond snake bites, their initial success suggests that artificial intelligence “can help democratize the discovery of therapies,” especially in the case of devastating neglected diseases, thanks to the “substantial” savings in money and resources.

More than two million people suffer snakebites every year, mostly in Africa, Asia and Latin America. The injected toxins can cause paralysis and bleeding. Despite the magnitude of the problem, current treatments employ a crude and often ineffective strategy developed more than a century ago: injecting horses with snake venom, extracting their blood and obtaining the specific antibodies generated. “Unfortunately, there is very little funding, either from academic institutions or from large pharmaceutical companies, to improve current treatments for snakebites,” laments Vázquez.

David Baker, born in Seattle 62 years ago, won half of last year’s Nobel Prize in Chemistry. The other half was divided between Demis Hassabis and John Jumper, two researchers at Google DeepMind who were instrumental in developing AlphaFold, a system that predicts the structure of proteins with unprecedented accuracy. To understand the complexity of the challenge, a water molecule has two hydrogen atoms and one oxygen atom, with the simple formula H₂O. The protein that allows us to breathe, the hemoglobin that makes our blood red, is C₂₉₅₂H₄₆₆₄N₈₁₂O₈₃₂S₈Fe₄.

Baker distances himself from the American multinational. “There is a big difference between my laboratory, which is totally open, we receive visitors from all over the world and we share information, and a company like Deepmind, which is totally closed,” the researcher stressed in an interview with EL PAÍS in 2023. “Being an open system gives you many more ideas. The free exchange of information benefits the advancement of science,” he said. Google DeepMind opened part of its systems after Baker shared his for free.

The Nobel Prize winner was very optimistic in a message sent to this newspaper on Tuesday. “I think designed proteins could help with many problem diseases!” exclaims Baker, director of the Institute for Protein Design at the University of Washington. “The advantage of design is that you can build in all the properties needed in the drug—this is very hard to do with other current methods of drug discovery,” he argues.

The Belgian biotechnologist Els Torreele helped found the Drugs for Neglected Diseases Initiative (DNDi), a nonprofit that seeks new treatments for conditions that affect more than 1 billion people. Torreele changed the face of pharmaceuticals in 2019 when she led Médecins Sans Frontières’ campaign for access to essential medicines. She and her colleagues wanted to prove that inventing a drug doesn’t cost €2.5 billion, as Big Pharma claimed. They succeeded. DNDi invested €55 million in developing fexinidazole, the first oral treatment for sleeping sickness, an infection transmitted by tsetse flies and caused by parasites that inflame the brain.

Torreele is skeptical about the promise of “democratization” that comes with artificial intelligence. “Democratizing drug discovery would assume that access to the big data and high-power computational tools would become widely available and affordable, such that anyone can use those in their labs. I doubt that this will be the case any time soon, given the costs of big data, data centers, and computational capacity (including carbon footprint),” says the biotechnologist.

The Belgian researcher does not believe that the generation of candidate drug molecules is the main bottleneck for drug development, at least in most diseases. Torreele believes that the real challenge – faced with a pharmaceutical industry that wants maximum benefits from its investments – is in the expensive clinical trials with thousands of people, to prove that an experimental treatment is safe and effective.

“With less (donor) finance overall for global health, also the non-profits like DNDi increasingly struggle to mobilize funds,” Torreele warns. “Of course it’s nice to have more drug candidates, but getting them developed into actual medicines in ways that deliver affordable and equitable access to patients when and where needed is the main bottleneck today, and I don’t see how AI is going to be helpful on that,” she warns.

Sign up for our weekly newsletter to get more English-language news coverage from EL PAÍS USA Edition