Google DeepMind unveils next-gen AlphaFold AI model to accelerate drug discovery

Over 91,000 users in India are users of AlphaFold Database, which provides free access to predicted protein structures (Image Credit: Google Deepmind)

Google DeepMind and its spinoff drug discovery firm Isomorphic Labs on May 8 rolled out a new version of AlphaFold. This artificial intelligence model can predict the structure and interactions of all life's molecules, including proteins and DNA.

The firms said that the new model, AlphaFold 3, built on the foundations of AlphaFold 2 had made a fundamental breakthrough in protein structure prediction in 2020, by predicting the 3D structure of a protein from its amino acid structure.

AlphaFold 3 expands beyond proteins to provide accurate predictions for protein interactions with other biomolecules in living cells - such as DNA, RNA, and small molecules.

"Biology is a dynamic system, and you have to understand how properties of biology emerge through the interactions between different molecules in the cell. You can think about AlphaFold 3 as our first big step towards that" said Google DeepMind CEO Demis Hassabis.

Google DeepMind said that more than 1.8 million researchers globally have used AlphaFold 2 to make discoveries in areas such as malaria vaccines, cancer treatments, and enzyme design. In India, over 91,000 researchers have used the AlphaFold Database which provides free access to over 200 million predicted protein structures, the firm said.

Read: India well-positioned to help shape future of AI: Google DeepMind's Jeff Dean

With AlphaFold 3, Google DeepMind claims that it has seen at least a 50 percent improvement compared with existing prediction methods, and more than doubled the prediction accuracy for some key categories of interaction.

"AlphaFold 3 brings the biological world into high definition. It allows scientists to see cellular systems in all their complexity, across structures, interactions and modifications" the firm said in a blog post.

Hassabis said a key advantage of AlphaFold 3 is modeling protein interaction with ligands. "'s incredibly important for drug discovery and it's something that we're pushing forward, especially Isomorphic Labs, towards a critical capability needed for discovery" he said.

Isomorphic Labs, a unit of Google parent Alphabet, said it is using AlphaFold 3 along with a suite of in-house AI models to accelerate and improve the success of drug design - by helping understand how to approach new disease targets and developing new ways to pursue existing ones that "were previously out of reach".

Read: DeepMind's GraphCast: The next generation of weather forecasting powered by Google AI

The firm is also collaborating with pharmaceutical companies to apply it to real-world drug design challenges with a goal to develop new "life-changing" treatments for patients.

Earlier this year, Isomorphic Labs had announced strategic tie-ups with Eli Lilly and Novartis, two of the world's leading pharmaceutical companies. At the time, the company had said that these partnerships have the potential to be worth nearly $3 billion, excluding any royalties that may result from future drug sales.

AlphaFold Server launch

Google DeepMind is also introducing a new AI tool, called AlphaFold Server, that will make predictions on how proteins interact with other molecules throughout the cell available to the scientific community and non-commercial researchers for free.

This will enable researchers to easily generate predictions, regardless of their access to computational resources or their expertise in machine learning.

The company said that researchers will be able to access AlphaFold 3’s capabilities to generate large and complex biological structures containing proteins, DNA, RNA, and ligands. They can also model chemical modifications for proteins and nucleic acids in a single platform.

"Experimental protein-structure prediction can take about the length of a PhD and cost hundreds of thousands of dollars. Our previous model, AlphaFold 2, has been used to predict hundreds of millions of structures, which would have taken hundreds of millions of researcher-years at the current rate of experimental structural biology" the firm said in a blogpost.

Google DeepMind said it will also expand its free AlphaFold education online course with the European Bioinformatics Institute (EMBL-EBI) and partnerships with various organisations to equip scientists with tools needed to accelerate adoption and research, including on underfunded areas such as neglected diseases and food security.