We are delighted to share that the inaugural Kavli Award Lecture for 2022 took place last week. Google DeepMind's CEO, Demis Hassabis, visited Kavli INsD to present his lecture and to receive his award. Kavli INsD Director Professor Carol Robinson introduced the speaker and described his outstanding career path. The event, which was also livestreamed via Zoom, continued with Demis Hassabis's lecture on "Using AI to Accelerate Scientific Discovery".
In his talk, Demis mentioned how they were able to solve the protein folding problem with AlphaFold. For 50 years, people have been working on protein folding, a problem first articulated by Nobel Prize winner Christian Anfinsen. In his Nobel acceptance speech in 1972, Anfinsen predicted that in theory, it should be possible to go from the one-dimensional sequence to the three-dimensional shape.
The normal way the 3D structure is determined requires painstaking experimental work. The rule of thumb is that it takes one Ph.D. student their entire Ph.D. (four or five years) to determine the structure of one protein. As a consequence, in the history of structural biology, only ~150,000 protein structures have been identified. DeepMind researchers used as their initial dataset these 150,000 proteins to train Alphafold. Remarkably, Alphafold can now take an amino acid sequence and predict its 3D structure in a matter of seconds with, on average, atomic accuracy.
Demis explained that predicting the 3D structure of a given protein was challenging because there are so many possibilities for amino acid sequences to fold. Templates, derived from known structures of similar proteins, are used to initiate computational searches for putative structures. Alternatively, Alphafold can use information about the biological evolution of amino acid sequences of the same protein family. The resulting Alphafold system is now widely used, including by Kavli INsD Professor Matt Higgins.
One of the most immediate applications of AlphaFold is drug discovery. When a researcher designs a new molecule or new compound, a potential drug, the key step is to figure out where in the protein that molecule will bind to either inhibit or block the protein’s function. This is just one part of the drug discovery process, but it's an important part. Additionally, a lot of diseases are currently thought to involve proteins that fold in the wrong way or misfold. Given the proposal that Alzheimer's disease might involve protein misfolding, for example of amyloid beta protein, it is possible that AlphaFold will help us to understand how amyloid fibrils form.
At the end of his lecture, a lively question and answer session took place with Kavli INsD researchers. Demis Hassabis then received his award from the Kavli INsD Director, Professor Dame Carol Robinson and the evening continued with a drinks reception followed by a celebratory dinner.