Google DeepMind has launched AlphaFold 3, an upgraded version of its AI model. This new model predicts the structure of not only proteins but also “all life’s molecules.”
AlphaFold 3‘s capabilities will aid researchers in various fields such as medicine, agriculture, materials science, and drug development. It will enable the testing of potential discoveries in these areas.
Previous versions of AlphaFold solely focused on predicting the structures of proteins. AlphaFold 3 extends its capabilities to model DNA, RNA, and smaller molecules known as ligands. This expansion enhances AlphaFold’s utility for scientific applications.
DeepMind announces a significant improvement in prediction accuracy with its new model. The new model, AlphaFold 3, demonstrates a 50 percent increase in accuracy compared to its predecessors.
DeepMind CEO Demis Hassabis highlights AlphaFold 2 as a milestone in structural biology, unlocking new research opportunities. AlphaFold 3 represents another step forward in utilizing AI for understanding and modeling biological processes.
AlphaFold 3 possesses a repository of molecular structures. Researchers provide a list of molecules for combination. AlphaFold 3 employs a diffusion technique to produce a 3D model of the resultant structure.
This diffusion method resembles the AI system utilized by image generators such as Stable Diffusion to assemble photos.
Isomorphic Labs, a drug discovery company founded by Hassabis, has integrated AlphaFold 3 into its internal projects. AlphaFold 3 has aided Isomorphic Labs in enhancing its comprehension of new disease targets.
DeepMind is offering AlphaFold Server, along with the model, to select researchers for free. AlphaFold Server, driven by AlphaFold 3, enables scientists to create biomolecular structure predictions without worrying about compute power.
The server is open for academic and noncommercial purposes, according to Hassabis. Isomorphic Labs is collaborating with pharmaceutical partners to utilize AlphaFold models in drug discovery initiatives.
Google is collaborating with the scientific community and policy leaders to ensure responsible deployment of AI models.
According to a paper from Google, certain biosecurity experts suggest that AI models could potentially decrease barriers for threat actors, allowing them, along with other technologies, to create pathogens and toxins that are more dangerous or easily transmissible.
The company collaborated with domain experts, biosecurity professionals, research specialists, and industry experts. They aimed to identify and address potential risks associated with AlphaFold 3 even prior to its launch.
Related Stories: