DeepEMhancer: a deep learning solution for cryo-EM volume post-processing

Abstract Cryo-EM maps are valuable sources of information for protein structure modeling. However, due to the loss contrast at high frequencies, they generally need be post-processed improve their interpretability. Most popular approaches, based on global B-factor correction, suffer from limitations. For instance, ignore heterogeneity in map local quality that reconstructions tend exhibit. Aiming overcome these problems, we present DeepEMhancer, a deep learning approach designed perform automatic post-processing cryo-EM maps. Trained dataset pairs experimental and sharpened using respective atomic models, DeepEMhancer has learned how post-process performing masking-like sharpening-like operations single step. was evaluated testing set 20 different maps, showing its ability reduce noise levels obtain more detailed versions Additionally, illustrated benefits SARS-CoV-2 RNA polymerase.