Google DeepMind’s new AI model can help with drug discovery

Google DeepMind has quietly open-sourced its groundbreaking artificial intelligence (AI) model that can predict the interactions between proteins and other molecules. The large language model, called AlphaFold 3, is the successor to AlphaFold 2, whose research led to large language model (LLM) creators Demis Hassabis and John Jumper being awarded the 2024 Nobel Prize in Chemistry. AlphaFold 3 takes things a step further with its ability to model the interactions of proteins with DNA, RNA and other smaller molecules, potentially leading to drug discovery.

Google DeepMind Open-Sources AlphaFold 3 AI model

Research into protein structures is one of the most important areas of interest in chemistry. Because the 3D shape and atomic details of proteins are targets for drugs, discovering new protein structures can often open up previously unexplored targets and mechanisms for medical intervention. Simply put, the better we understand protein structures, the more effective drugs can be against various conditions, diseases and autoimmune diseases.

Although Google DeepMind has not made any announcement about releasing the AlphaFold 3 AI model, it has determined the source code and model weight available on GitHub. However, this is only available for academic and research purposes. The source code is freely available under a Creative Commons license, but the weights are only accessible after direct permission from Google for academic use.

It is believed that if the AI ​​model can correctly highlight how proteins interact with DNA, RNA and other smaller molecules, researchers will be able to speed up the production of new synthetic drugs.

Researchers will also be able to automate work that could have taken them years without any evidence of success. AlphaFold 3 comes three years after the release of AlphaFold 2 in 2021 studythe lead author emphasized that drug discovery could become much easier using the AI ​​model.

The AlphaFold 3 has been trained on a vast amount of research material and datasets on protein structures and their interactions with other molecules. By understanding the context and logic of protein structures, the LLM can predict how certain target zones will respond when coming into contact with certain molecules.