Google has been busy over the past year developing its Gemini chatbot to compete with ChatGPT, a multifunctional AI assistant designed to assist with work tasks and personal digital chores. However, in a quieter manner, the company has also been enhancing its AlphaFold software, developed by Google’s DeepMind AI unit, which predicts the 3D structure of proteins. The software has been upgraded to model other molecules of biological significance, such as DNA and the interactions between antibodies and disease organisms.
DeepMind CEO Demis Hassabis emphasized the significance of these upgrades for drug discovery, as AlphaFold 3 can accurately predict how molecules will interact with each other, from large entities like DNA to smaller molecules like metal ions. The software, developed by Google DeepMind and Isomorphic labs, will be available for outside researchers to access for free via the cloud.
Previously, predicting protein structures required laborious processes using electron microscopes and x-ray crystallography. However, with the advancement of deep learning, as demonstrated by AlphaFold, the shape of proteins can now be accurately predicted from their constituent amino acids.
In 2020, AlphaFold 2 produced impressive results, leading to excitement in the field of molecular biology. Subsequently, in 2021, Google released an open source version of AlphaFold along with 350,000 predicted protein structures, and later expanded to over 2 million protein structures. This development signifies a significant advancement in the field of AI in biotech and drug discovery.
