Deep learning machine AlphaFoldwhich was created by the Google AI research lab Deep Mindis already changing our understanding of the molecular biology underlying health and disease.
One half Nobel Prize 2024 in Chemistry went to David Baker from the University of Washington, USA, and the other half awarded jointly Demis Hassabis AND John M. Jumpereach from London-based Google DeepMind.
If you have not heard of AlphaFold, it could also be hard to understand how essential it is becoming to researchers. However, as a beta tester of the software, I had the opportunity to see firsthand how this technology could reveal the molecular structures of assorted proteins in a matter of minutes. It would take researchers months and even years to separate these structures in laboratory experiments.
This technology could pave the way for latest, revolutionary therapies and medicines. But first, it’s essential to grasp what AlphaFold does.
Proteins are created as a results of a series molecular “beads”created from the choice of the human body 20 different amino acids. These beads form a long chain that folds into mechanical shape that is crucial for the functioning of the protein.
Their sequence is set by DNA. And although DNA testing signifies that we know the order of the beads that make up most proteins, predicting how the chain will assemble into each “3D machine” has all the time been a challenge.
Illustrations: Niklas Elmehed © Nobel Prize Awareness Campaign, CC BY-NC-SA
These protein structures form the basis of all biology. Scientists study them in the same way that you just might take a clock apart to grasp how it works. Understand the parts and put the whole together: it’s the same with the human body.
Proteins are tiny, and there are huge numbers of them in each of them our 30 trillion cells. This meant that for a long time the only strategy to determine their shape was through laborious experimental methods – research that might take years.
Throughout my profession, like many other scientists, I have been engages in such activities. Every time we solve a protein structure, we put it into a global database called Protein Data Bankwhich everyone can use for free.
AlphaFold was trained on these structures, most of which were used X-ray crystallography. In this method, proteins are tested in hundreds of various chemical states, with changes in temperature, density and pH. Scientists use a microscope to find out the conditions under which each protein assembles into a specific formation. They are then exposed to X-rays to find out the spatial arrangement of all the atoms in the protein.
After being trained in these designs, AlphaFold can now just do that predict the structure of a protein at speeds that were previously not possible.
I began at the starting of my profession, in the late Nineties, developing protein structures using the magnetic properties of their nuclei. I did this using a technology called nuclear magnetic resonance (NMR), which uses a huge magnet just like an MRI scanner. This method began to fall out of favor resulting from some technical limitations, but that is the way it is now is experiencing a rebirth thanks AlphaFold.
NMR is one among the few techniques that may study molecules in motion, quite than holding them stationary in a crystal or on an electron microscope grid.
Rywka Isaacson, Provided by creator (no reuse)
An addictive experience
In March 2024, DeepMind researchers asked me to check the beta version of AlphaFold3, the latest incarnation of the software that was near release at the time.
I’ve never been a gamer, but I got a taste of the addictive experience because once I gained access, all I desired to do was spend hours trying out molecular combos. In addition to lightning-fast speed, this new edition introduced the ability to include larger and more diverse molecules, including DNA and metals, in addition to the ability to switch amino acids to mimic chemical signaling in cells.
Our laboratory at King’s College London used X-ray crystallography predict the structure formed by two bacterial proteins which can be loosely involved hospital superbugs once they interact. Previous incarnations of AlphaFold predicted individual components but could never solve the problem accurately – and yet the new edition solved the problem the first time.
Understanding the moving parts and dynamics of proteins is the next frontier now that we are able to predict the static shapes of proteins using AlphaFold. Proteins are available in a huge number of sizes and shapes. They may be rigid or flexible, or product of rigorously structured units connected by flexible loops.
Dynamics are essential for protein function. As one other Nobel Prize winner – said Richard Feynman: “Everything that living things do can be understood in terms of the oscillations and vibrations of atoms.”
Another great feature of magnetic resonance techniques is the ability to exactly measure the distances between atoms. So, after some rigorously designed experiments, AlphaFold’s results may be verified in the laboratory.
In other cases, the results are still inconclusive. This is a work in progress between experimental structural biologists like my team and computational scientists.
The recognition that comes with a Nobel Prize will only spur the pursuit of understanding the entire molecular machinery and hopefully change the landscape when it involves drugs, vaccines and human health.
