Link to bioRxiv paper: http://biorxiv.org/cgi/content/short/2020.08.28.271981v1?rss=1 Authors: He, J., Huang, S.-Y. Abstract: Motivation and Results: Advances in microscopy instruments and image processing algorithms have led to an increasing number of cryo-EM maps. However, building accurate models for the EM maps at 3-5 [A] resolution remains challenging and time-consuming. Here, we present a fully automatic de novo structure determination method using a deep learning-based framework, named as DeepMM, which automatically builds atomic-accuracy all-atom models from cryo-EM maps at near-atomic resolution. In our method, the main-chain and C positions as well as their amino acid and secondary structure types are predicted in the EM map using Densely Connected Convolutional Networks. DeepMM was extensively validated on 40 simulated maps at 5 [A] resolution and 30 experimental maps at 2.6-4.8 [A] resolution as well as an EMDB-wide data set of 2931 experimental maps at 2.6-4.9 [A] resolution. DeepMM built correct models for >60% of the cases, and it outperformed existing state-of-the-art algorithms including RosettaES, MAINMAST, and Phenix. Availability: http://huanglab.phys.hust.edu.cn/DeepMM/ Note: This manuscript has been submitted to Nature Communications on June 17, 2020. Copy rights belong to original authors. Visit the link for more info