Google DeepMind introduced an improved version of its AI model that predicts not only the structure of proteins, but also the structure of DNA and RNA. The work of the new AlphaFold 3 model will help researchers in medicine, agriculture, materials science and drug development test potential discoveries.
Previous versions of AlphaFold only predicted protein structures. AlphaFold 3 goes beyond this and can model DNA, RNA and smaller molecules called ligands, expanding the model’s capabilities for scientific use.
A DeepMind representative said the new model shows a 50 percent improvement in prediction accuracy compared to previous models. “With AlphaFold 2, it was a major milestone in structural biology and has opened up all kinds of amazing research,” said DeepMind CEO Demis Hassabis. “AlphaFold 3 is a step forward in using AI to understand and model biology.”
AlphaFold 3 has a library of molecular structures. Researchers enter a list of molecules they want to combine, then AlphaFold 3 uses a diffusion method to create a 3D model of the new structure. Diffusion is the same type of AI system that AI image generators like Stable Diffusion use to compose photos.
Along with the model, DeepMind also makes the AlphaFold Server research platform available to some researchers for free. According to Hassabis, the server is available for academic and non-commercial use.
Google noted that it worked with biosecurity, research and industry experts to prevent the risks associated with AlphaFold 3 even before it was launched.
Source: The Verge