- Paul Rincon
- Scientific Affairs Editor, BBC News Network
Artificial intelligence is used to predict the structure of almost every protein produced by the human body.
This major scientific research development can accelerate the development of new drugs for the treatment of diseases, and can also be used in many other places.
Protein is an integral part of all biological organisms, and every cell in our body is full of protein.
Understanding the structure and shape of proteins is crucial to the advancement of medical research, but until now, we have only studied the structure of a few proteins.
Researchers use an artificial intelligence software called “AlphaFold” to predict the structure of 350,000 proteins in the human body and other organisms.
The composition of human proteins is determined by the genome in our body, that is, the genetic material DNA in the human cell nucleus, which can determine about 20,000 proteins.
“AlphaFold” artificial intelligence software was developed by Google’s DeepMind artificial intelligence company. The company’s co-founder and CEO Demis Hassabis said that this is the most complete and accurate image of human proteins so far. .
“We believe this represents the greatest contribution that artificial intelligence has made to the progress of scientific knowledge.”
He said that this is the best example of AI’s contribution to society, and there will be more motivating developments in the future.
Proteins are composed of polypeptide chains formed by the combination of amino acids in a certain order. They are folded into various unique three-dimensional shapes in countless ways. The three-dimensional shape of the protein determines its function in the human body.
The 350,000 protein structures predicted by AlphaFold include the 20,000 protein composition of the human body, as well as the proteins of the so-called “model organisms”, which are organisms used for scientific research in the laboratory, such as the large intestine. Bacillus, yeast, fruit flies and white mice.
DeepMind researchers and the European Molecular Biology Laboratory (European Molecular Biology Laboratory) team published research reports in the journal Nature.
The test results show that AlphaFold artificial intelligence software can accurately predict 58% of the amino acid structure of human proteins, and 35.7% of the protein structure is more accurate, which is twice the result of laboratory research.
Traditional methods for studying protein structure include X-ray crystallography, low-temperature electron microscopy, etc., but it is very difficult to study with traditional methods. Structural biologist John McGeehan of the University of Portsmouth ) The professor said that traditional methods require a lot of money and resources.
Therefore, traditional methods to study protein structure are usually part of targeted scientific research. Before this, there has never been any research project that can systematically study the structure of all proteins in the human body.
In fact, before that, only 17% of protein structures had been studied in the laboratory.
Professor Mark Jihan said that the speed of artificial intelligence is very fast. It took 6 months to study the structure of a protein from the beginning, and now it only takes a few minutes, which is impossible with traditional methods.
“We sent 7 sequences to the DeepMind research team at the beginning, two of which have already been developed in the laboratory, so we can compare and test their accuracy when they send the results back.”
“We saw that the structure predicted by AlphaFold was exactly the same as the structure obtained by our laboratory, which is frankly shocking.”
Professor Edith Heard of the European Molecular Biology Laboratory said that protein is the most basic part of an organism. Artificial intelligence has studied the structure of proteins, which has completely changed how we understand the operation of life.
The results of this development have a wide range of applications in the future, including the development of new drugs and treatments that can be used for various diseases, the design or modification of future crops that can withstand climate change, and the creation of enzymes that can digest plastics.
Professor Mark Jihan’s research team has begun to use AlphaFold data to develop enzymes that can digest plastics faster. He said that the protein structure prediction provided by this artificial intelligence software is impossible in the laboratory, which has accelerated their project for several years. time.
DeepMind has cooperated with the European Molecular Biology Laboratory to open AlphaFold’s coding and protein structure predictions to global scientific research groups.
DeepMind’s Hassabis plans to substantially expand the database to include the structures of more than 100 million proteins known to the scientific community.