Artificial Intelligence Cracks a 50-Year-Old Problem in Biology
By admin - December 20, 2020

The artificial intelligence of DeepMind, which works on protein processing, has met the 50-year-old challenge of biology. This progress will make it much easier for researchers to develop drugs and study diseases. It can even lead to a complete coup in medicine, scientists point out to us. Many of them admit in the media that they did not think they would witness such progress in their lives.

However, before we talk directly about moving forward, it is worth noting that DeepMind has already shown itself many times. He proved that artificial intelligence can handle both complex tasks as well as the superhuman.

Demi Hasabis, the public face and co-founder of DeepMind, has always stressed that these successes were just the beginning and that they were moving towards a bigger goal. Hasabis said that artificial intelligence will definitely help us in cognition of the universe.

Most recently, DeepMind and the organizers of the Critical Assessment of Protein Structure Prediction (CASP) literally confirmed Hasabis’ predictions. DeepMind’s latest version of AlphaFold, a deep machine learning system that can work to determine the structure of proteins, has accomplished what was impossible for 50 years.

“Artificial intelligence has not solved such a serious problem so far,” said John Molt, who heads CASP.

A protein is made up of a ribbon of amino acids that is wrapped around itself in many intricate strands. It is this structure that determines the protein. And the study of this entanglement is essential for a good understanding of the mechanisms of life. For example, the development of coronavirus vaccines focuses on spike proteins. The way a coronavirus enters a human cell depends on the spike protein and its forms. However, in addition to the spike protein, there are tens of thousands of different types of proteins in humans.

In CASP, AlphaFold predicted the structure of dozens of proteins with an error limit of 0.16 nanometers (atom size). A similar calculation method is much faster than other hitherto existing and first-time accuracy of laboratory techniques such as cryo-electron microscopy, nuclear magnetic resonance and X-ray crystallography. Most importantly, the techniques listed are expensive and slow: they usually involve a lot of effort and hundreds of thousands of dollars. Work on this process can take years, and all this to determine the shape of each protein. AlphaFold can do the same in a few days.

Yes, we are dealing with really serious progress! This progress will take medicine to a new level.

In the long run, predicting protein structure will help us develop synthetic proteins such as enzymes that break down waste and produce biofuels. In addition, researchers are exploring ways to introduce synthetic proteins that will increase crop yields and make plants more fertile.

“This is a very important achievement,” said Mohamed Alkuraishi, a systemic biologist at Columbia University, who has developed his own program for predicting protein structure. “This is something I just did not expect, I did not think it would happen so fast. That fact must be taken into account. “

Alkurai thought researchers would need at least 10 years to get such results from AlphaFold. “These measurements are literally on the physical edge,” he said.

Protein structure is very difficult to identify. For most proteins, researchers have an amino acid sequence in the ribbon, but not for the forms in which they are folded. When it comes to folding and tangling, we are already talking about the astronomical number of shapes. Researchers have been trying to solve this challenge since the 1970s, when Christian Anfinsen received the Nobel Prize and proved that structure is defined in sequence.

This may also mean that the predicted structure is an operating alternative configuration, detected in the laboratory, in the range of natural variations.

The mention requires that AlphaFold rely on the collaborative work of hundreds of researchers. DeepMind encompassed numerous fields that brought together teams of biologists, physicists, and computer scientists.

Now researchers are trying to figure out exactly how AlphaFold works. Scientists say that once all this becomes even more understandable, the world will change. People will use this technique for all sorts of different things, and these things will be something we can’t even imagine right now.