AlphaFold an artificial intelligence tool has managed to predict the structure of 350 000 human proteins, 44% of all proteins in the body. Before this breakthrough, scientists knew only about 170 000 of them.
Artificial Intelligence
Scientists have managed to develop a database containing the protein structures of 350 000 human proteins and 20 other organisms, including yeasts. This achievement was made possible thanks to an artificial intelligence tool called AlphaFold, whose algorithm was trained to recognize the amino acids that make up the structure of proteins. To do this, the researchers used data from 170 000 protein structures they already knew. Their work was published in the journal Nature.
Read Also: An Artificial Intelligence Is Now Able to Predict Cardiovascular Diseases Through The Eyes
Understanding amino acids to predict protein structure
To understand the significance of this discovery, we must first clearly define the terms. First, proteins are the main building blocks of all cells in the human body. They instruct cells to carry out their functions. The set of proteins in an organism is called a proteome. Its composition changes according to environmental conditions, both inside and outside the cells. The study of the proteome – and thus of the thousands of proteins that make up the body – is interesting because it aims to improve our understanding of the mechanisms of life and our organisms.
However, in order to analyze proteins, we need to study their amino acids. These are the molecules that give proteins their shape by folding them. Each folded protein has a specific conformation that gives it a different shape from other proteins. These folds, therefore, create a wide variety of proteins. The challenge for researchers is to be able to predict these folds in order to predict the possible structure of proteins. Before AlphaFold, only 170 000 protein structures were known.
Read Also: An Artificial Intelligence Capable of Translating Thoughts Into Words Is Now a Reality
A better understanding of proteins for drug design
“AlphaFold was able to do in a weekend what it took us months and years to do,” said John McGeehan in a comment on the study. I feel like we just jumped at least a year from where we were yesterday. Once the AI algorithm is developed, the analysis is very fast. On the other hand, this AI tool also revealed many disordered human proteins, which means that the algorithm was not able to predict their structure.
The authors conclude that these proteins do not have a unique but variable shape. The primary aim of this list of 350 000 proteins is to improve our basic knowledge of human, plant, and animal biology. For example, it can be used by all scientists to create new drugs, because some proteins are resistant or involved in infection. Currently, a team of researchers is using AlphaFold to analyze how coronavirus proteins attach to human cells.
References
Highly accurate protein structure prediction for the human proteome
FEEDBACK: