Google Shuts Down Its Nobel Prize-Winning AlphaFold Project
Credit: deepmind.google
Google’s artificial intelligence (AI) division, DeepMind, has disbanded the AlphaFold team, which created the protein structure prediction tool of the same name.
The AlphaFold artificial intelligence system accurately predicted three-dimensional protein structures from amino acid sequences in minutes, not years. It is now used to accelerate drug discovery, vaccine development, and the analysis of structural changes in proteins associated with neurodegenerative diseases, including Alzheimer’s and Parkinson’s.
DeepMind began developing AlphaFold in 2018. In 2020, the program celebrated solving a 50-year-old problem facing humanity—the “protein folding problem,” which involved solving the question of how amino acids fold themselves into complex three-dimensional shapes. These shapes determine the biological role of proteins. Over the past half-century, scientists have determined the structures of approximately 170,000 proteins, using methods such as X-ray diffraction and nuclear magnetic resonance. The AlphaFold developers used the results of this work to train the AI.
In 2021, the journal Nature published a paper describing the AlphaFold methodology and predicting the structure of the entire human proteome—the complete set of proteins the human body can produce. DeepMind opened the AlphaFold protein structure database, providing researchers with free access to over 200 million AI-predicted protein structures. In 2024, DeepMind head Demis Hassabis and John Jumper, the engineer who pioneered the project, were awarded the Nobel Prize for their work on AlphaFold.
Google DeepMind’s vice president of research, Pushmeet Kohli, told the FT that these changes are due to a shift in the company’s strategy. He said that instead of focusing on specific areas, the company is now focusing on developing Gemini-based systems that can help scientists across a range of fields.
