Share to Facebook Share to Twitter Share to Linkedin InnovationRx is your weekly digest of healthcare news. To get it in your inbox, subscribe here . A protein folding prediction in response to a small molecule generated by AlphaFold 3.
T oday, Deepmind announced the launch of AlphaFold 3 , the next generation of its AI software for structural biology, which has enabled thousands of researchers to predict the complicated chemical interactions involved with proteins. One major difference in the new iteration of the program: instead of just predicting protein folding , it predicts the interactions of other biological molecules as well including DNA, RNA and small molecules. The company published the accuracy of its predictions in the journal Nature .
In addition to the new capabilities–which even includes the ability to make predictions based on epigenetic changes to DNA–the company has also launched AlphaFold Server , a web service that enables researchers to generate biochemical models without having to install the system itself from open source code , as was the case with previous iterations. Something else new is that AlphaFold now works with diffusion models – which are popularly known for being used by AI systems that generate images from text prompts – in place of some of its previous structural models. “The response to AlphaFold 2 was more than I could have ever imagined in terms of the kind of creativity and what the research community has done with it,” De.
