Improving cryo-EM maps by resolution-dependent and heterogeneity-aware deep learning

Cryo-electron microscopy (cryo-EM) has emerged as a leading technology for determining the structures of biological macromolecules. However, map quality issues such as noise and loss of contrast hinder accurate map interpretation. Traditional and deep learning-based post-processing methods offer improvements but face limitations particularly in handling map heterogeneity. Here, we present EMReady2, a substantial evolution of our previously developed deep learning-based methodology, towards a generalist model for improving cryo-EM maps. EM-Ready2 introduces a local resolution-dependent training strategy, allowing it to effectively address map heterogeneity. Additionally, EMReady2 extends its applicability to cryo-EM maps with nucleic acids, intermediate-resolution maps, and cryo-electron tomography (cryo-ET) maps. Extensive evaluations on diverse test sets of 94 maps at 2.0–10.0 Å resolutions demonstrate that EMReady2 is a universally applicable post-processing model for improving EM density maps in terms of both quality and interpretability, and significantly outperforms existing methods. EM-Ready2 is freely available at http://huanglab.phys.hust.edu.cn/EMReady2/ .