From: allin
Google’s DeepMind team members Demis Hassabis and John Jumper received the Nobel Prize in Chemistry for their work on AlphaFold [03:06:08]. This recognition highlights a significant convergence of hard sciences and computer science [03:34:34].
What is AlphaFold and Protein Folding?
AlphaFold addresses a long-standing challenge in biochemistry: predicting or visualizing the three-dimensional structure of proteins [03:09:50]. Proteins are formed by long chains of amino acids that collapse onto themselves in a specific three-dimensional shape, which determines their physical and structural function [03:10:03]. Understanding this “protein folding” was extremely difficult, often relying on methods like X-ray imaging systems or complex computational models [03:10:24].
Historically, efforts to tackle this problem with computing date back decades [03:00:03]. For example, Stanford had a “Folding@Home” distributed computing project in the early 2000s that used idle CPU cycles from participants’ computers to model protein folding [03:11:36]. The AlphaFold project at DeepMind ultimately solved this complex problem [03:45:00].
Impact and Applications
The ability to predict the 3D structure of proteins from their amino acid sequence is incredibly important [03:05:05]. It unlocks the human capacity to create biomolecules designed for specific functions [03:16:19]. This includes:
- Binding cancer cells [03:22:20]
- Breaking apart pollutants and plastics [03:22:24]
- Creating entirely new molecules [03:28:28]
- Designing micromotors from proteins, as demonstrated by David Baker at the University of Washington [03:31:31]
Since AlphaFold’s initial publication, literally dozens of companies have been founded, and billions of dollars in capital have been invested into companies focused on creating new drugs and industrial biotech applications using this protein modeling capability [03:32:09].
Future Developments and Broader Implications
DeepMind has since released AlphaFold 3 [03:50:50] and launched a drug discovery company called Isomorphic Labs, which predicts molecules for specific target indications using AlphaFold models [03:53:54].
The recognition of AlphaFold’s creators, alongside Jeffrey Hinton’s work in training deep neural networks [03:30:28], highlights a significant trend: the convergence of hard sciences and computer science [03:33:34]. This era sees AI advancements and computers being used to enhance our understanding of the natural sciences [03:53:01]. Google has also made the protein structure for 200 million proteins available for free through DeepMind [03:41:20].