Topaz-Denoise

Topaz-Denoise denoises cryo-electron microscopy (cryo-EM) micrographs and cryo-electron tomography (cryoET) datasets to increase signal-to-noise ratio for improved particle detection and structural analysis.

Key Features:

• Deep Learning-Based Denoising: Utilizes convolutional neural networks trained on thousands of micrographs collected under varied imaging conditions to model cryo-EM image formation and remove noise.
• Generalization Across Datasets: Employs pretrained models that can denoise new datasets without requiring additional retraining.
• Improved Micrograph Interpretability: Enhances signal-to-noise ratio to aid identification of particles and particle orientations for downstream processing.
• Facilitation of Low Dose Collection: Enables denoising of low electron-dose images, reducing radiation damage and accelerating data acquisition.
• 3D CryoET Denoising Model: Includes a general 3D denoising model applicable to cryo-electron tomography volumes.

Scientific Applications:

• Clustered protocadherin structure determination: Enabled resolution of the first 3D single-particle closed and partially open structures of clustered protocadherin.
• Low-dose structural analysis: Facilitates low-dose cryoEM and cryoET data collection for more efficient and less radiation-damaging structural studies.

Methodology:

Convolutional neural networks are trained on a comprehensive dataset reflecting a wide range of imaging conditions, and the trained models are applied to denoise micrographs and tomograms, including a general 3D model for cryoET.