Urdnet: A Cryo-EM Particle Automatic Picking Method

Cryo-Electron Microscopy (Cryo-EM) images are characterized by the low signal-to-noise ratio, contrast, serious background noise, more impurities, less data, difficult data labeling, simpler image semantics, and relatively fixed structure, while U-Net obtains resolution when downsampling rate information to complete object category recognition, high-resolution during upsampling precise segmentation positioning, fills in underlying through skip connection improve accuracy of segmentation, has advantages biological processing like Cryo-EM image. This article proposes A based residual intensive neural network (Urdnet), which combines point-level pixel-level tags, used accurately automatically locate particles from cryo-electron microscopy images, solve bottleneck that cryo-EM Single-particle macromolecule reconstruction requires tens thousands picked particles. The 80S ribosome, HCN1 channel TcdA1 toxin subunits, other public protein datasets have been trained tested on Urdnet. experimental results show Urdnet could reach same excellent particle picking performances as mainstream methods RELION, DeepPicker, acquire 3D structure with higher resolution.