an advance on the structure of proteins

Recently published in the prestigious review Science *, the results obtained by a group of New York and Californian researchers bear witness to the current effervescence in laboratories: in a few months, these scientists have been able to create proteins playing an active role in the organization’s regulation processes. A job which, by the sole means of natural selection, would have asked about 500 million years to occur!

Proteins, due to their particular chain composition of animated acids, take a three -dimensional shape, necessary for their activity. This configuration gives them a capacity for biological action in cells: producing energy, transmitting information, facilitating communication between cells, transforming other molecules, etc. The Ba-Ba of cell life, in short, made possible by these small molecular machines that are proteins.

Until now, understanding this three -dimensional structure was an almost impossible mission. It took twelve years to the biologist John Kendrew (1917-1997) to decipher, the first, the structure of myoglobin, determining in the storage and transport of oxygen within muscle cells. A Benedictine work which earned him the Nobel Prize in chemistry in 1962.

It was not until 1994 that things accelerated: CASP, a new kind scientific competition, was organized between teams of researchers to develop innovative software in order to accelerate this important work of decryption. Its organizers also offer neophytes to participate in this laborious work of analysis through a video game created for the occasion.

A giant leap

One of the three winners of the Nobel Prize in Chemistry 2024, Demis Hassabis, then a young prodigy of chess and video games, was part of the adventure. In 2010, he created the Deepmind company, bought since by the giant Google and now famous, since it has become one of the spearheads of progress in artificial intelligence (IA).

In 2018, Demis Hassabis presented with his colleague John Jumper, also a Colauréat du Nobel, Alphafold, an analysis tool specifically devoted to the prediction of protein structures. Over the improvements made to it, this generative AI software upsets the field of research.

Where, over six decades, the researchers had deciphered the structure of nearly 150,000 different proteins, these new tools only took a few months to dissect nearly 200 million. Almost all of the known proteins existing in nature.

What already facilitate current medical research in many areas: development of vaccines (malaria, etc.) and new cancer drugs, autoimmune diseases, diabetes, fight against antibiotic resistance, etc. The future accelerates.

* Published on January 16, 2025. Vol 387, p. 850 to 858. Available on the scientific review website.