AI-Designed Drugs At Medical Stores Near You? What AI's Big Breakthrough In Biology With AlphaFold 3 Means

A new artificial intelligence (AI) model developed by Google DeepMind and Isomorphic Labs promises to transform our understanding of the biological world and elevate drug discovery to unprecedented heights.

The model is called AlphaFold 3, as it’s the third iteration in this line of AI models, and can predict the structure and interactions of all life’s molecules with accuracy unlike anything seen before for a non-experimental method. AlphaFold was the first non-experimental method to be able to make a protein structure prediction.

For decades, scientists have used experimental methods like X-ray crystallography, nuclear magnetic resonance (NMR) spectroscopy, and cryoelectron microscopy to predict the three-dimensional (3D) structure of proteins from the 1D sequence of amino acids. An amino acid sequence — there are 20 different types of amino acid — carries all the information necessary to reliably fold into a 3D protein.

While the experimental methods work fine, and remain the gold standard, AI has now entered the mix to help scientists save on time and money. “Experimental protein-structure prediction can take about the length of a PhD and cost hundreds of thousands of dollars,” DeepMind says.

Not with AlphaFold 2. The model has already been used to predict hundreds of millions of structures that “would have taken hundreds of millions of researcher-years at the current rate of experimental structural biology.”

Knowledge of the 3D structure of proteins sheds light on how proteins go about their work, how they interact with other molecules, and how they can be targeted for drug design.

The latest model, AlphaFold 3, is capable of joint structure prediction of complexes including proteins, nucleic acids, small molecules, ions, and modified residues. So, it goes beyond proteins to cover a broad range of biomolecules.

Speaking to Bloomberg, Google DeepMind chief executive and co-founder Demis Hassabis said, “Of course, biology, as we know, is a dynamic system. Really, all the emergent properties of biology in life are due to interactions between different molecules and different structures. So, that’s what AlphaFold 3 is about, as a first step in that direction.” 

As detailed in a paper in research journal Nature, with the latest model, scientists have observed much higher accuracy on protein-ligand interactions (ligands are small molecules) than the advanced tools in use currently, on protein-nucleic acid interactions than nucleic-acid-specific predictors, and significantly so with antibody-antigen prediction as compared to AlphaFold 2.

The paper reports an improvement in prediction of 50 per cent over other methods for the interactions of proteins with other molecule types, while “for some important categories of interaction we have doubled prediction accuracy,” its scientist-authors say.

“Together these results show that high accuracy modelling across biomolecular space is possible within a single unified deep learning framework,” according to the paper.

However, unlike with AlphaFold 2, access to the upgrade comes with restrictions. AlphaFold 3 is limited to non-commercial use through a DeepMind website. Scientists, for instance, will not be able to run their own version of AlphaFold 3, as Ewen Callaway writes in Nature.

“Instead, researchers will have access to an ‘AlphaFold3 server’, on which they can input their protein sequence of choice, alongside a selection of accessory molecules,” he adds.

The AlphaFold Server is described as “the most accurate tool in the world for predicting how proteins interact with other molecules throughout the cell.” The research tool will help scientists who are looking to come up with novel hypotheses to test in the laboratory.

Google DeepMind has put up a demonstration of the AlphaFold Server on their YouTube channel. The video takes users through the steps on how to model a protein with ions, bound to ribonucleic acid or RNA.

RNA is one of the three major biological macromolecules, alongside deoxyribonucleic acid (DNA) and proteins, which are essential for all known forms of life.

As for AlphaFold 3’s potential for real-world application in drug design, DeepMind collaborator Isomorphic Labs is in touch with pharmaceutical companies. The protein structure predictor, it is thought, will help resolve real-world drug design challenges and lead to the creation of new life-changing treatments for patients.

“I would be expecting maybe in the next couple of years the first AI-designed drugs in the clinic,” Hassabis told Bloomberg.

AlphaFold 2 already has a rich resume. Apparently, millions of researchers across the world have used it to make discoveries in areas including malaria vaccines, cancer treatments, and enzyme design.

“This leap (AlphaFold 3) could unlock more transformative science, from developing biorenewable materials and more resilient crops, to accelerating drug design and genomics research,” Google said.

“If you’d ask me the number one thing AI could do for humanity, it would be to solve, you know, hundreds of terrible diseases. I can’t imagine a better use case for AI,” Hassabis says. AlphaFold 3 is a step in that direction.