by AlphaFold 3, unveiled to the world on May 9, is the latest version of an algorithm designed to predict the structures of proteins – vital molecules used by all life – from the “instruction code” in their building blocks.
One of the biggest problems in biology is predicting protein structures and the way they interact with other molecules. However, AI developer Google DeepMind has gone some way to solving it in the last few years. This new version of the AI system has improved functionality and accuracy over its predecessors.
Like the next release in a video game franchise, structural biologists – and more recently – chemists are waiting impatiently to see what they can do. It is widely understood that DNA is the instruction book of a living organism but, inside our cells, proteins are the molecules that actually do most of the work.
They are proteins that enable our cells to perceive the outside world, integrate information from different signals, make new molecules within the cell, decide to grow or stop growing.
They are also proteins that enable the body to distinguish between foreign invaders (bacteria, viruses) and itself. And the targets of most of the drugs you or I take to treat disease are proteins.
Lego protein
Why is protein structure important? Proteins are large molecules containing thousands of atoms in very specific orders. The order of these atoms, and how they are organized in 3D space, is crucial for a protein to be able to fulfill its biological function.
This same 3D arrangement also determines the way a drug molecule binds to its protein target and treats disease.
Imagine you have a Lego set where the bricks are not based on cuboids, but can be any shape. In order to fit two bricks together in this row, each brick will need to fit snugly against each other without any holes. But that’s not enough – both bricks will have the right combination of bumps and holes to keep the bricks in place.
Designing a new drug molecule is like playing with this new Lego set. Someone has already built a giant model (the protein target found in our cells), and the drug discovery chemist’s job is to use their tools to assemble a bunch of bricks that bind to a specific part of the protein and – in terms biological – stop it from performing its normal function.
So what does AlphaFold do? Based on knowing exactly what atoms are in any protein, how these atoms have changed differently in different species, and what other protein structures are similar, AlphaFold is very good at predicting the 3D structure of any protein .
AlphaFold 3, the latest iteration, has expanded capabilities to model nucleic acids, for example, pieces of DNA. It can also change the shape of proteins with chemical groups that can turn the protein on or off, or with sugar molecules. This gives scientists more than just a bigger and more colorful Lego set to play with. It means they can develop more detailed models to read and correct the genetic code and cellular control mechanisms.
This is important for understanding disease processes at a molecular level and for developing drugs that target proteins whose biological role is controlling which genes are turned on or off. The new version of AlphaFold also predicts antibodies with better accuracy than previous versions.
Antibodies are important proteins in biology in their own right, and are a vital part of the immune system. They are also used as biological drugs such as trastuzumab, for breast cancer, and infliximab, for diseases such as inflammatory bowel disease and rheumatoid arthritis.
The latest version of AlphaFold can predict the structure of proteins bound to small molecules such as drugs. Drug discovery chemists can already predict the way a potential drug binds to its protein target if the target’s 3D structure is identified through experiments. The downside is that this process can take months or even years.
Predicting how potential drugs and protein targets bind to each other is used to help decide which potential drugs to synthesize and test in the laboratory. Not only can AlphaFold 3 predict drug binding in the absence of an experimentally identified protein structure but, during testing, it outperformed existing software predictions, even if the target structure and the drug binding site is known.
These new capabilities add AlphaFold 3 to the arsenal of tools used to discover new therapeutic drugs. More accurate predictions will allow better decisions to be made about which drugs might be tested in the lab (and which are unlikely to be effective).
Time and money
This saves both time and money. AlphaFold 3 also provides the opportunity to make predictions about drug binding to modified forms of the protein target that are biologically relevant but difficult – or impossible – to do using existing software. Examples of this are proteins modified by chemical groups such as phosphates or sugars.
Of course, as with any potential new drug, extensive experimental testing for safety and efficacy – including human volunteers – is always required before it is approved as a licensed medicine.
AlphaFold 3 has several limitations. Like its predecessors, it is poor at predicting the behavior of protein regions that do not have a fixed or ordered structure. It is poor at predicting multiple conformations of a protein (which may change shape due to drug binding or as part of its normal biology) and cannot predict protein dynamics.
It can also make somewhat embarrassing chemical mistakes, such as placing atoms on top of each other (physically plausible), and substituting mirror images for certain structural details (biologically or chemically impossible).
A more significant limitation is that the code will not be available – at least for now – so it will have to be used on the DeepMind server on a non-commercial basis only. Although this will not affect many academic users, it will limit the enthusiasm of expert modellers, biotechnologists and many applications in drug discovery.
Despite this, the release of AlphaFold 3 looks set to spark a new wave of creativity in drug discovery and structural biology more broadly – and we’re already looking forward to AlphaFold 4.
This article from The Conversation is republished under a Creative Commons license. Read the original article.
The authors do not work for, consult with, or own shares in, or receive funding from, any company or organization that would benefit from this article, and have disclosed no relevant affiliations beyond their academic appointment.
