Deepmind’s AlphaFold program successfully predicts a protein’s 3D structure
DeepMind, the AI developed by Google. has made what scientists are calling a “gargantuan leap” in solving biology’s greatest question. To say scientists are excited by the discovery would be an understatement. Andrei Lupas, an evolutionary biologist at the Max Planck Institute for Developmental Biology in Tübingen, Germany told reporters:
“It’s a game changer. This will change medicine. It will change research. It will change bioengineering. It will change everything.”
The AI program, called AlphaFold, uses off-the-shelf AI components such as HHblits, PSI-BLAST, TensorFlow, and Sonnet to train with more than 43 GB (compressed) worth of data. It competed against 100 teams in the every-other-year Critical Assessment of Structure Prediction (CASP) challenge. The results were announced today – AlphaFold crushed the competition with its newfound ability to predict protein’s 3D structure from their amino-acid sequence.
Why protein structure prediction is important
In molecular biology, “structure if function” guides the experimental process and proteins, the building blocks of all life forms, drive what happens inside cells. Thus the protein’s structure is critical to areas of science concerning living things (e.g. biology, medicine). The structure of a protein largely determines the protein’s function. However, protein structures can be difficult to determine experimentally. Today, biologists conduct laboratory experiments to map out the amino acids that make up a protein structure. Traditional structure prediction uses X-ray crystallography or cryo-electron microscopy, but even then the results are often sketchy.
The CASP experiment (they don’t like to call it a “competition”) was created to accelerate the development of protein structure predictions. The event, which takes place over several months, challenges teams to predict the structure of proteins that have been experimentally created but not yet revealed to the public. Target proteins or portions of proteins called domains — about 100 in total — are released on a regular basis and teams have several weeks to submit their structure predictions. Teams use various techniques to solve the problem. In this year’s competition, about two-thirds of AlphaFold’s structure predictions were indistinguishable from the traditional “gold standard” methods.
How AlphaFold’s new prediction logic will impact scientific research
Why is the discovery such a big deal? Researchers are calling this a “revolution in biology”. The discovery vastly accelerates efforts to understand the building blocks of cells. It lets scientists study living things in a new and novel manner. This new ability for instance, enables quicker and more advanced drug discovery.
One researcher said,
“I think it’s fair to say this will be very disruptive to the protein structure prediction field. I suspect many will leave the field as the core problem has arguably been solved. It’s a breakthrough of the first order, certainly one of the most significant scientific results of my lifetime.”
Indeed, AlphaFold may be the first AI to work out a grand scientific challenge that humans were unable to solve.
