Topaz-Denoise: general deep denoising models for cryoEM and cryoET

This paper introduces Topaz-Denoise, a deep learning method for enhancing the signal-to-noise ratio (SNR) of cryoEM images and cryoET tomograms. Trained on a diverse dataset of micrographs, Topaz-Denoise learns to denoise new datasets without further training, improving micrograph interpretability and enabling the solving of 3D structures. The method also facilitates low-dose data collection, accelerating the data acquisition process. A general 3D denoising model for cryoET is also presented. Topaz-Denoise is publicly available and integrated into several cryoEM software packages.

Tristan Bepler, Kotaro Kelley, Alex J. Noble, Bonnie Berger