An artificial intelligence (AI) system called “AlphaFold,” developed by the British company “DeepMind,” a subsidiary of Google , has solved one of the greatest mysteries in biology that has puzzled scientists for decades. This artificial intelligence (AI) system can quickly and accurately predict how proteins will fold, what three-dimensional shape they will take, and what their function will be .
This is an achievement that has been described as a milestone by the scientific world, and is expected to help better understand various diseases and accelerate the development of drugs for various diseases, including COVID-19. Other diseases that will be focused on are malaria and leishmaniasis.
Demis Hassabis, founder and CEO of DeepMind, said that this achievement is an exciting moment for the scientific community, noting that these algorithms are now becoming mature enough and powerful enough to be applied to truly challenging scientific problems.
"I was really amazed when I saw it. I never expected to see it in my lifetime," said structural biologist Dr. John Moult of the University of Maryland, one of the organizers of the international scientific competition CASP (Critical Assessment of protein Structure Prediction), which began in 2004 and is held every two years.
Tens of thousands of different proteins play a vital role in the body's health and disease, as cells are made of proteins, the complex shape of which plays an important role. For example, the coronavirus that causes COVID-19 enters human cells because a prominent spike protein on it structurally matches a receptor protein on human cells.
Furthermore, the shape of each protein is determined by the sequence of 20 different amino acids, which form chains to make proteins (for this discovery, the Nobel Prize was awarded to Christian Anfinsen in 1972). It is easy for biologists to "read" this sequence, as it is determined by the DNA that encodes it, but it is almost impossible - despite half a century of efforts - to predict the shape of a protein based on its tens or hundreds of amino acids.
Until now, scientists have resorted to experimental techniquessuch as X-ray crystallography, cryo-electron microscopy and MMR spectroscopy, which are very difficult, take months or years and do not always work correctly. But the new deep learning system AlphaFold, which began testing in 2018 and has since been significantly improved, can now associate any amino acid sequence with a different protein structure. However, this new system also has some weaknesses. For example, it does not do so well with proteins whose structure is affected by interactions with other proteins.
Evolutionary biologist Andrei Lupas of the Max Planck Institute for Developmental Biology in Germany said that for the next few years, researchers will continue to test the accuracy of AlphaFold's predictions themselves, but after that they will rely solely on artificial intelligence (AI) calculations. He said the real revolution will come when scientists can use computers to predict how proteins interact with other molecules. He stressed that this will completely change the face of medicine.
It is worth noting that the AlphaFold algorithm has already correctly predicted the shapes of several SARS-CoV-2 coronavirus proteins. In the future, it may be able to predict which of the thousands of existing drugs bind correctly to these proteins, without the need for scientists to conduct expensive and complex experiments.
The creators of AlphaFold have not revealed many details about their system, but they have said that they will soon publish a scientific paper on it. Finally, DeepMind has not yet clarified exactly how scientists around the world will be able to use the new system, but it has pledged to ensure broad access.
