DeepMind’s new AI can predict structure of ‘all life’s molecules’

  • AlphaFold 3 has shown a 50% improvement in prediction accuracy compared to previous versions, expanding its capabilities to model DNA, RNA, and ligands.
  • DeepMind CEO Demis Hassabis highlighted the significant utility of AlphaFold 3 in various scientific research fields and its potential for enabling groundbreaking discoveries.
  • DeepMind is providing the AlphaFold Server research platform, powered by AlphaFold 3, to some researchers for generating biomolecular structure predictions at no cost, enhancing research accessibility and collaboration.

Google DeepMind is unveiling AlphaFold 3, an improved AI model that predicts the structure of proteins and “all life’s molecules”. This advancement will aid researchers in medicine, agriculture, materials science, and drug development.

AlphaFold 3 shows 50% improvement from previous ones

Previous iterations of AlphaFold were limited to predicting protein structures. However, the latest version, AlphaFold 3, has transcended this boundary by including the ability to model DNA, RNA, and smaller molecules known as ligands, thus broadening the model’s utility for scientific research.

In a media briefing, DeepMind CEO Demis Hassabis expressed that AlphaFold 2 was a significant milestone for structural biology, enabling groundbreaking research opportunities. AlphaFold 3 represents a progression in utilising AI to comprehend and simulate biological processes.

DeepMind reports a 50% enhancement in prediction accuracy with the introduction of the new model compared to its predecessors.

Also read: Google incorporates cybersecurity into AI plan

Also read: Meta’s AI unveils full image creation for advertisers

The new model assists research in multiple fields

DeepMind says Isomorphic Labs, a drug discovery company founded by Hassabis, has been using AlphaFold 3 for internal projects. So far, the model helped Isomorphic Labs improve its comprehension of novel disease targets.

In addition to the model, DeepMind is providing the AlphaFold Server research platform to some researchers at no cost. The server, powered by AlphaFold 3, enables scientists to generate biomolecular structure predictions regardless of their computational resources. Hassabis says that the server is accessible for academic and non-commercial purposes, yet Isomorphic Labs is collaborating with pharmaceutical partners to utilise AlphaFold models in drug discovery initiatives.

The company stated that it collaborated with domain experts, biosecurity specialists, and industry professionals to identify potential risks associated with AlphaFold 3 even before its official launch.


Crystal Feng

Crystal Feng is an intern news reporter at Blue Tech Wave dedicated in tech trends. She is studying Chinese-English translation at Beijing International Studies University. Send tips to

Related Posts

Leave a Reply

Your email address will not be published. Required fields are marked *