An AI tool for predicting protein shapes could be transformative for medicine, but it challenges science’s need for evidence

<spanklasse=Christoph Burgstedt / Shutterstock” src=”–/YXBwaWQ9aGlnaGxhbmRlcjt3PTk2MDtoPTYwNg–/ dbc05a8c221dfd7″ data-src= “–/YXBwaWQ9aGlnaGxhbmRlcjt3PTk2MDtoPTYwNg–/ c221dfd7″/>

An advanced algorithm developed by Google DeepMind has gone some way to unraveling one of the greatest unsolved mysteries in biology. AlphaFold aims to predict the 3D structures of proteins based on the ‘instruction code’ in their building blocks. The latest upgrade was recently released. The latest upgrade was recently released.

Proteins are essential parts of living organisms and participate in virtually every process in cells. But their shapes are often complex and difficult to visualize. Being able to predict their 3D structures therefore provides insight into the processes in living things, including humans.

This offers new possibilities for making medicines to treat diseases. This in turn opens up new possibilities in what is called molecular medicine. This is where scientists strive to identify the causes of diseases on a molecular scale and also develop treatments to correct them on a molecular level.

The first version of DeepMind’s AI tool was unveiled in 2018. The latest version, released this year, is AlphaFold3. A global competition to evaluate new ways to predict the structures of proteins, the Critical Assessment of Structure Prediction (Casp) has been held biennially since 1994. In 2020, the Casp competition was allowed to test AlphaFold2 and was very impressed. Since then, researchers have eagerly anticipated each new incarnation of the algorithm.

However, as a graduate student I was once reprimanded for using AlphaFold2 in some of my courses. This was because it was only considered a predictive tool. In other words, how could anyone know if what was predicted matched the real protein without experimental verification?

This is a legitimate point. The field of experimental molecular biology has undergone its own revolution in the past decade with strong advances in a microscope technique called cryo-electron microscopy (cryo-EM), which uses frozen samples and soft electron beams to capture the structures of biomolecules in high resolution . .

The advantage of AI tools such as AlphaFold is that it can elucidate protein structures much faster (within minutes) and virtually free of charge. Results are more easily available and accessible online worldwide. They can also predict the structure of proteins that are notoriously difficult to verify experimentally, such as membrane proteins.

However, AlphaFold2 is not designed to address something called the quaternary structure of proteins, where multiple protein subunits form a larger protein. This includes a dynamic visualization of how different units of the protein molecule are folded. And some researchers reported that it sometimes seemed to have difficulty predicting structural elements of proteins known as coils.


When my professor contacted me in May to pass on the news that AlphaFold3 had been released, my first question was about its ability to predict quaternary structures. Did it work? Were we now able to make the quantum leap to predicting a complete structure? Initial reports suggest the answers to these questions are positive.

Experimental methods are slower. And when they can capture the 3D structure of molecules, it’s more like looking at a statue—a snapshot of the protein—rather than seeing how it moves and interacts to perform actions in the body. In other words, we want a movie instead of a photo.

Experimental methods have also traditionally struggled with membrane proteins – key molecules attached to or associated with the membranes of cells. These are often crucial for understanding and treating many of the worst diseases.

This is where AlphaFold3 could really change the landscape. If it succeeds in predicting quaternary structures at a level equal to or greater than that of experimental methods such as crystallography, cryo-EM and others, and can visualize membrane proteins better than the competition, then we will indeed see a quantum leap make in our development. race towards true molecular medicine.

AlphaFold3 can only be accessed from a DeepMind server, but is easy to use. Researchers can easily extract their results from the array within minutes. The other promise of AlphaFold3 is further disruption. DeepMind is not alone in its ambitions to master the problem of protein folding. As the next Casp competition approaches, there are others who want to win the race. For example, Liam McGuffin and his team at the University of Reading are making progress in assessing the quality and predicting the stoichiometry of protein complexes. Stoichiometry refers to the ratios in which elements or chemical compounds react with each other.

Not all scientists in this field pursue this goal in the same way. Others are trying to solve similar challenges in terms of the quality of the 3D models or specific barriers such as those of membrane proteins. The competition in this area was great.

However, experimental methods are not going away anytime soon, nor should they. The progress of cryo-EM is commendable, and X-ray crystallography still gives us the best resolution for biomolecules. The European XFEL laser in Germany could be the next breakthrough. These technologies will only continue to improve.

My biggest question as we explore this new field is whether our human instinct to give in until we have absolute proof will fail with AlphaFold. If this new technology can produce results comparable to, or greater than, experimental verification, will we be willing to accept it? If we can do that, its speed and accuracy could have a major impact on areas like drug development.

For the first time, with AlphaFold3 we may have overcome the most important hurdle in the protein prediction revolution. What will we think of this new world? And what medicine can we make with it?

This article is republished from The Conversation under a Creative Commons license. Read the original article.

The conversationThe conversation

The conversation

Sam McKee does not work for, consult with, own shares in, or receive funding from any company or organization that would benefit from this article, and has disclosed no relevant affiliations beyond their academic appointment.

Leave a Comment