Several companies have developed open-source protein structure prediction tools based on AlphaFold3, including Chinese technology giant Baidu, TikTok developer ByteDance, and San Francisco start-up Chai Discovery. However, these models, like AlphaFold3, are not licensed for commercial applications such as drug discovery. Other teams are working on versions of AlphaFold3 without such restrictions, with a fully open-source model, OpenFold3, expected to be available by the end of the year.
Key takeaways:
- Google DeepMind has released the software code for AlphaFold3, an AI tool that can predict protein structures, for non-commercial applications.
- AlphaFold3 can model proteins in concert with other molecules, which was not possible with its predecessors. This could potentially allow scientists to predict how proteins behave in the presence of potential drugs.
- Several companies have developed their own open-source protein structure prediction tools based on AlphaFold3, including technology giants Baidu and ByteDance, and San Francisco start-up Chai Discovery.
- DeepMind's decision to release the code for AlphaFold3 has been met with both praise and criticism, with some scientists arguing that withholding the code in the first place undermined reproducibility.