DeepMind makes AlphaFold3 open source
Nov. 13, 2024.
2 mins. read.
13 Interactions
Google DeepMind has made AlphaFold3, an AI model for predicting protein structures, available as open-source for non-commercial use.
Google DeepMind has made AlphaFold3, an AI model for predicting protein structures, available as open-source for non-commercial use, Nature News reports.
AlphaFold predicts how proteins fold into their three-dimensional shapes, which is crucial for understanding their function.
This decision came after some controversy over the initial release of the model without its source code. AlphaFold3 is an upgrade from its earlier versions, like AlphaFold2, because it can predict not only protein structures but also how proteins interact with other molecules, which is vital for drug discovery.
Previously, DeepMind made AlphaFold3 accessible only through a web server, which limited the types of predictions users could make, particularly regarding how proteins interact with drugs.
Scientists criticized this method because it hindered reproducibility – the idea that scientific results should be able to be replicated by others to verify their validity.
Scientists argued that for research to be trustworthy, the underlying code should be available. This led DeepMind to promise the release of the code within six months.
The importance of open source code for AI
Now, with the code released, researchers can explore how proteins might interact with potential drugs, aiding in drug discovery.
However, the model weights, which are the learned parameters that allow the AI to make predictions, are still only available to academic users upon request.
The push for openness in AI models stems from the scientific community’s desire for transparency and verifiability in research. This openness allows for further innovation and application of the technology in unexpected ways. In fact, researchers used previous versions of AlphaFold in creative ways like designing new proteins for medical targets.
“We’re very excited to see what people do with this,” says AlphaFold project leaderJohn Jumper as reported by Nature News. Jumper won the 2024 Nobel Prize in Chemistry with CEO Demis Hassabis for their work on AlphaFold2. “People will use it in weird ways,” he adds. “Sometimes it will fail and sometimes it will succeed.”
Let us know your thoughts! Sign up for a Mindplex account now, join our Telegram, or follow us on Twitter.
0 Comments
0 thoughts on “DeepMind makes AlphaFold3 open source”