LIVERPOOL, ENGLAND - Artificial intelligence could be about to supercharge the way the human race finds new drugs and treats illness.

New software unveiled by Google’s AI division Deepmind this month holds the potential to change the way we design new drugs, and could allow us to target diseases more effectively, says Prof Daniel Rigden of the University of Liverpool’s Department of Biochemistry, Cell and Systems Biology.

Rigden says that AlphaFold3 allows scientists a ‘window’ into biological processes, by predicting how molecules will interact. He says: "The program may be particularly important in developing new drugs since it can accurately predict how a drug target - typically a protein - binds to a drug, typically a small organic molecule.’

How does it work?

It predicts the ways that different proteins interact with each other and (uniquely) with DNA and RNA, Rigden explains.

AlphaFold3 is better than other methods of doing so, and the new version performs better than its predecessor, AlphaFold2, and can perform a wider range of predictive tasks, he says.

A Google spokesperson told Yahoo News: "AlphaFold 3 is a revolutionary AI model that can predict the structure and interactions of all of life’s molecules. It helps us to visualise how the components of our cells communicate with each other - so we can build a better understanding of the vital processes in the human body, other species and even plants.

"AlphaFold 3 is a jump in accuracy to Google DeepMind’s previous AlphaFold 2 model, and significantly improves accuracy in key biomolecule classes."


What could it achieve?

Ridgen says that Alphafold3’s importance comes from its ability to accurately predict the structures of molecules and the way these interact with each other.

Rigden says that Alphafold3 will also help in "literally any area of biological research".

Google suggests that the software could assist with anything from discovering effective new treatments and tackling diseases, to tackling the viruses affecting plants to help with food security challenges. It could also aid our understanding how our DNA is read, copied and repaired to keep us thriving as a species.

Google Deepmind’s own Isomorphic Labs is already using AlphaFold3 to find new drugs.

A Google spokesperson said: "Combined with other AI tools developed at Isomorphic Labs, we believe there is potential to accelerate (by reducing the time required for drug design), improve (by allowing us to better understand and characterise disease targets) and ultimately transform (by equipping us to go after entirely new targets to benefit distinct patient populations) drug discovery.

"This work accelerates our understanding of the molecular machines powering the human body - and has the potential to unlock a world of possibility in areas like plant immunity, drug design, understanding biorenewable materials, inspiring genomics research and more."

Why will it make a difference?

Determining protein structures by experimentation (which is still the ‘gold standard’ of what AlphaFold3 does) can cost up to $100,000.

Being able to predict structures in software could therefore save vast amounts of time and money.

A database of 200 million protein structure predictions from AlphaFold2 is already widely used.

A Google spokesperson said: "The platform’s state-of-the-art predictions cover nearly every type of molecule in the Protein Data Bank, which houses all the known biomolecular structures figured out by painstaking experiments.

"In combination with AlphaFold 3, Google DeepMind also launched the AlphaFold Server. The AlphaFold Server is a web-based tool that allows scientists to generate predictions of these cellular interactions with a few clicks of a button, completely free of charge, to propel forward this critical scientific research."

Dr Nicole Wheeler, Birmingham Fellow, Institute of Microbiology and Infection, University of Birmingham, said: ““Unlike general trends toward adding more parameters at each new iteration of a model to achieve higher performance, AlphaFold 3 actually reduces the number of steps and computational complexity of their model. This is an important advance as the US government start placing reporting requirements on models that use a lot of compute, and people become more conscious of the carbon footprint of training models with increasing levels of complexity.

"Physically producing and testing biological designs is a big bottleneck in biotechnology at the moment, so this is very encouraging for the prospect of rapidly prototyping biological parts for new applications, ranging from medicines, to food, to environmental applications.”