DeepMind’s next-generation protein structure predictor AlphaFold 3 launched

The new AlphaFold 3 AI system for protein structure prediction has been released by Google DeepMind and its isomorphic spinout, which is built around the system. According to the companies, the innovative system now shows at least a 50 percent improvement compared to existing prediction methods and, for some important interaction categories, the prediction accuracy has doubled.

AlphaFold 3 aims to go beyond proteins and cover a broad spectrum of biomolecules. The system and some results are described in a paper published in Natureand the main author is Google DeepMind Josh Abramson.

The system uses a novel end-to-end deep neural network trained to produce protein structures from amino acid sequences, multiple sequence alignments, and homologous proteins. In their paper, the researchers write, the AlphaFold 3 model has “a substantially updated diffusion-based architecture, which is capable of predicting the joint structure of complexes including proteins, nucleic acids, small molecules, ions, and modified residues.”

The isomorphic team reports“To harness the potential of AlphaFold 3 for drug design, we at Isomorphic Labs are already collaborating with pharmaceutical companies to apply it to real-world drug design challenges and ultimately develop new life-changing treatments.” patients.”

Since its launch about two years ago, Isomorphic has announced two agreements worth $3 billion with Eli Lilly and Novartis to apply AI to discover new drugs.

The new model is based on the foundations of AlphaFold 2, which in 2020 made a Fundamental breakthrough in protein structure prediction. The companies report that so far, millions of researchers Globally, AlphaFold 2 has been used to make discoveries in areas including malaria vaccines, cancer treatments and enzyme design. The system has been cited more than 20,000 times and has received numerous accolades, most recently the Breakthrough Prize in Life Sciences.

Given an input list of molecules, AlphaFold 3 generates their joint 3D structure, revealing how they all fit together. Models large biomolecules such as proteins, DNA and RNA, as well as small molecules. Additionally, AlphaFold 3 can model chemical modifications of these molecules that control healthy cell function.

‍Its developers say AlphaFold 3’s capabilities come from its next-generation architecture and training that now covers all the molecules of life. The core of the model is an improved version of the evoformer module—a deep learning architecture that underpinned AlphaFold 2’s performance.

After processing the inputs, AlphaFold 3 assembles its predictions using a diffusion network, similar to those found in AI imagers. The diffusion process begins with a cloud of atoms and, over many steps, converges to its final, more precise molecular structure.

‍The companies report that AlphaFold 3’s predictions of molecular interactions exceed the accuracy of all existing systems.

They write in their report: “The new AlphaFold model demonstrates significantly improved accuracy over many previous specialized tools: much higher accuracy in protein-ligand interactions than state-of-the-art docking tools, much higher accuracy in protein-nucleic acid interactions than specific predictors nucleic acid and significantly higher. Antigen-antibody prediction accuracy than AlphaFold-Multimer v2.3.”

Scientists can access most of AlphaFold 3’s capabilities, for free, through the newly released AlphaFold Server research tool. ‍