In 2021, A.I research Laboratory Deep Mind announced the development of its first neural network for digital biology, AlphaFold. The model was able to accurately predict the three-dimensional structure of proteins, which determines the functions these molecules perform. “We are just floating bags of water,” says Pushmeet Kohli, vice president of research at DeepMind. “What sets us apart are proteins, the building blocks of life. It’s how they interact with each other that makes the magic of life happen.
AlphaFold was recognized by the journal Science as a Breakthrough of the Year 2021. In 2022 it was the most cited research paper on artificial intelligence“People were on [protein structures] for many decades and we have not been able to make that much progress,” says Kohli. “Then artificial intelligence came along.” DeepMind also released the file AlphaFold protein structure database— which contained the protein structures of almost every organism whose genome had been sequenced — making it freely available to scientists around the world.
More than 1.7 million researchers in 190 countries they have used it for research ranging from designing plastic-eating enzymes to developing more effective malaria vaccines. A quarter of the research involving AlphaFold was devoted to understanding cancer, Covid-19 and neurodegenerative diseases such as Parkinson’s and Alzheimer’s. Last year, DeepMind released the next generation of AlphaFold, which extended its structure prediction algorithm to biomolecules such as nucleic acids and ligands.
“It democratized research,” Kohli says. “Scientists working in a developing country on a neglected tropical disease did not have access to the funding needed to calculate the protein structure. Now, with one click, they can go to the AlphaFold database and get free forecasts.” For example, one of DeepMind’s first partners, the Drugs for Neglected Diseases Initiative, used AlphaFold to develop a drug for diseases that affect millions of people, such as sleeping sickness, Chagas disease, and leishmaniasis, and yet are the subject of relatively little research.
DeepMind’s latest discovery is called AlphaMissense. This model categorizes so-called missense mutations – genetic changes that can produce different amino acids at specific positions in proteins. Such mutations can alter the function of the protein itself, and AlphaMissense assigns a score to the probability that the mutation is pathogenic or benign. “Understanding and predicting these effects is key to discovering rare genetic diseases,” Kohli says. The algorithm, which was released last year, classified about 89 percent of all possible human missenses. Previously, only 0.1 percent of all possible variants had been clinically classified by researchers.
“This is just the beginning,” says Kohli. He believes that artificial intelligence could ultimately lead to the creation of a virtual cell that could dramatically speed up biomedical research by enabling biology to be studied in silico rather than in real laboratories. “Thanks to artificial intelligence and machine learning, we finally have the tools to understand this highly sophisticated system we call life.”
The article was published in the July-August 2024 issue WIRED Magazine UK.