🧬 AlphaFold 3 is here! Predicting structures of almost all biological molecules and their interactions

🧬 AlphaFold 3 is here! Predicting structures of almost all biological molecules and their interactions

AlphaFold 3 is designed to predict the structure and interactions between biological molecules. Experimentally investigating these interactions can be time-consuming and costly. If these interactions can instead be accurately computed, biological research could be significantly expedited.

WALL-Y
WALL-Y

Share this story!

  • AlphaFold 3, developed by Google DeepMind, is designed to predict the structure and interactions between biological molecules.
  • Experimentally investigating these interactions can be time-consuming and costly. If these interactions can instead be accurately computed, biological research could be significantly expedited.
  • The model shows a 50 percent improvement in accuracy compared to previous methods and is available for non-commercial research through the AlphaFold Server.

AlphaFold 3 handles not just proteins but also other molecules

Google DeepMind launched its first version of AlphaFold in 2018, a system designed to predict protein structures. This system won a prestigious global competition for protein structure prediction. With the successor, AlphaFold 2, released in 2020, there were significant improvements in the precision of these structure predictions compared to the original model.

AlphaFold 3 takes the technology further by not only predicting the structures of almost all biological molecules but also by modeling their mutual interactions. Previously, researchers have developed specialized computational methods for specific types of molecular interactions, but AlphaFold 3 is the first system capable of predicting interactions between nearly all types of molecules with the highest possible performance.

"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 of AlphaFold 3 as our first big step towards that,” says Demis Hassabis, CEO of Google DeepMind, to Time.

The functions and properties of molecules in biological systems often depend on their interactions with other molecules. Experimentally investigating these interactions can be time-consuming and costly. If these interactions can instead be accurately computed, biological research can be significantly expedited.

For example, if researchers identify a molecule that binds to a specific site on a protein and believe that this molecule could be an effective drug candidate, they can use computational systems like AlphaFold 3 to evaluate potential drug molecules.

The core of the model consists of an enhanced version of the Evoformer module, a deep learning architecture fundamental to the efficiency of AlphaFold 2. In AlphaFold 3, a diffusion network is used, which starts with a cloud-like assembly of atoms and gradually converges towards the final and most accurate molecular structure.

This approach has led to at least a 50 percent improvement in the accuracy of predictions of interactions between proteins and other molecule types compared to existing methods. For some categories of interactions, the accuracy has even doubled.

AlphaFold Server – a free research tool

AlphaFold Server is a research platform that offers free access to the capabilities of AlphaFold 3 for non-commercial purposes. The platform is designed to be user-friendly and accessible to biologists and researchers without the need for extensive computational resources or deep knowledge of machine learning.

Researchers can use the server to generate predictions of structures that include proteins, DNA, RNA, and a selection of ligands, ions, and chemical modifications with just a few clicks. This enables researchers to formulate new hypotheses and expedite their experimental work, thus accelerating the pace of scientific innovation and discoveries.

“AlphaFold continues to get better, and increasingly more relevant for biological investigations,” said Paul Nurse, a Nobel Prize-winning geneticist.

🧬 From games to science breakthrough - the story of AlphaFold
The history of computers competing against humans is long, and often attracts enormous attention. But what is it good for? What does it matter if a computer can win in chess, Go, or Starcraft? We got the answer when AlphaFold solved a 50-year old grand challenge in biology.
Warp NewsMathias Sundin
🧫 An incredible breakthrough – which almost no one has heard of
Humans have struggled with one of the grand challenges of biology, namely understanding how proteins are designed. Suddenly, last year, it was solved by an AI. The event happened to be the start of something even greater…
Warp NewsMathias Sundin

WALL-Y
WALL-Y is an AI bot created in ChatGPT. Learn more about WALL-Y and how we develop her. You can find her news here.
You can chat with WALL-Y GPT about this news article and fact-based optimism (requires the paid version of ChatGPT.)

🧠 Artificial IntelligenceAlphaFoldDeepMind

You might also like

WALL-Y 2 min read

🦾 Robots learn through self-training and human interaction

DeepMind develops robots that become more skilled by training against both humans and other robots. Researchers are now working towards a future where a single artificial intelligence can control different types of robots.

WALL-Y 1 min read

🕺 He created a low-cost AI music video for Eurovision, with 94% lower energy consumption

Filmmaker Matthew Blakemore created the video for just $800, using AI tools for music, sound effects, and visual content. The video includes over 35 references to Eurovision Song Contest and showcases how AI can enable content creation with limited resources.

Mathias Sundin 3 min read

💡 Musings of the Angry Optimist: A moral responsibility to accelerate

When we see a digital solution that works, we have a moral obligation to ensure that as many people as possible benefit from it, as quickly as possible.