Deep-learning with synthetic data enables automated picking of cryo-EM particle images of biological macromolecules

• DOI: 10.1093/bioinformatics/btz728
• Publication Date: 2019-09-26T11:28:48Z
• Authors: Ruijie Yao; Jiaqiang Qian; Qiang Huang
• Keywords: Cryo-Electron Microscopy; Particle (ecology)

Single-particle cryo-electron microscopy (cryo-EM) has become a powerful technique for determining 3D structures of biological macromolecules at near-atomic resolution. However, this approach requires picking huge numbers macromolecular particle images from thousands low-contrast, high-noisy electron micrographs. Although machine-learning methods were developed to get rid bottleneck, it still lacks universal that could automatically the noisy cryo-EM particles various macromolecules.Here, we present deep-learning segmentation model employs fully convolutional networks trained with synthetic data known structures, called PARSED (PARticle SEgmentation Detector). Without using any experimental information, segment in whole micrograph time, enabling faster than previous template/feature-matching and particle-classification methods. Applications six large public datasets clearly validated its ability pick sizes. Thus, our method break particle-picking bottleneck single-particle analysis, thereby accelerates high-resolution structure determination by cryo-EM.The package user manual noncommercial use are available as Supplementary Material (in compressed file: parsed_v1.zip).Supplementary Bioinformatics online.