Researchers present RoseTTAFold - a neural network approach for protein modeling with accuracies near those of what DeepMind's AlphaFold2 has achieved, according to a new study. What's more, the powerful new tool's code, as well as a public server, are being made freely available to the scientific community. Proteins are the building blocks of life and their functions - central to nearly all biological processes - are directly related to their three-dimensional shape. However, determining a protein's 3D shape from its amino acid sequence alone has been a longstanding challenge. Recently, at the CASP14 protein structure prediction assessment conference, London-based DeepMind presented their new artificial intelligence (AI) network called AlphaFold2, which demonstrated the remarkable ability to predict the shape of proteins at a level of accuracy comparable to what is possible using expensive and time-consuming experimental methods such as X-ray crystallography and cryo-electron microscopy (cryo-EM). AlphaFold2's performance raised the question of whether such accuracy could be achieved in other systems developed by those outside of a world-leading deep learning company. Here, Minkyung Baek, David Baker and colleagues present RoseTTAFold - an AI network that produces structure predictions with accuracies approaching those of DeepMind's AlphaFold2, but using only a fraction of the computational processing power and time. Building upon the reported methodological advances of AlphaFold2 and their own CASP14 approach, Baek et al. developed a 3-track network that integrates and simultaneously processes 1-dimensional (1D) protein sequence information, 2D distance map information and 3D atomic coordinate information. The accurate predictions from the RoseTTAFold network enable the rapid solution of challenging X-ray crystallography and cryo-EM structure determination problems and provide insights into the functions of proteins of currently unknown structures, according to the authors.