Supervised Cryo-EM Tomography Denoising Enabled by Physics-Based Image Simulation

Data Codes

The data and source codes for the paper [Y. Kim and Y. Yang, "Supervised Cryo-EM Tomography Denoising Enabled by Physics-Based Image Simulation", Appl. Phys. Lett. 128 (2026).] are posted below.

Overview

Cryo-electron tomography enables three-dimensional visualization of biomolecules in their native aqueous state but suffers from low signal-to-noise ratio (SNR) due to electron dose limitations. Subtomogram averaging improves resolution but averages out structural heterogeneity, while self-supervised denoising risks over-smoothing fine details. Supervised learning has not been widely applied to cryo-ET because of the challenge of generating realistic training data. Here, we present a physics-informed supervised denoising framework based on a dedicated simulation pipeline. Microscope parameters, detector noise, and electron dose were extracted from experimental data, and clean images were generated through multislice calculations that solve the quantum-mechanical propagation of the electron wave function. Paired with noisy counterparts, these images were used to train a U-Net convolutional neural network. Applied to simulated and experimental tilt-series of liposomes, the method improved SNR by up to 132% and enhanced bilayer contrast without averaging, enabling recovery of fine structural details under low-dose conditions.

System Requirements

1. All software dependencies and operating systems (including version numbers)Software: MATLAB R2021a - academic use, TEMSIM, Python 3.6, Pytorch, Jupyter NotebookOperating systems: CentOS Linux 7

2. Versions the software has been tested onMATLAB R2021a academic usePython 3.6Jupyter Notebook 6.5.7

3. Any required non-standard hardwareThere are no non-standard hardware

Repositary Contents

1. Experimental Cryo-ET Tilt-Series

Folder: 1_Experimental_Cryo-ET_tilt_series

This folder contains cryo-ET tilt-series data of POPC liposome pairs acquired experimentally.It includes the tilt-series of Pair 1, recorded with a total electron dose of 340 e*(Angstrom^(-2)), and Pair 2, recorded with a total dose of 30 e*(Angstrom^(-2)).

Subfolder: 1_1_Pair1_340This subfolder contains the tilt-series of Pair 1, acquired with a total electron dose of 340 e*(Angstrom^(-2)).Exp_Pair1_tiltseries_raw.mat: the raw tilt-series of Pair 1.Exp_Pair1_tiltseries_registration.mat: the tilt-series of Pair 1 after image registration.Exp_Pair1_tiltseries_after_processing.mat: the post-processed tilt-series of the cropped liposome pair (Pair 1), including image registration.Exp_Pair1_tiltseries_angles.mat: the tilt angles of the Pair 1 tilt-series.

Subfolder: 1_2_Pair2_30This subfolder contains the tilt-series of Pair 2, acquired with a total electron dose of 30 e*(Angstrom^(-2)).Exp_Pair2_tiltseries_raw.mat: the raw tilt-series of Pair 2.Exp_Pair2_tiltseries_registration.mat: the tilt-series of Pair 2 after image registration.Exp_Pair2_tiltseries_after_processing.mat: the post-processed tilt-series of the cropped liposome pair (Pair 2), including image registration.Exp_Pair2_tiltseries_angles.mat: the tilt angles of the Pair 2 tilt-series.

2. Experimental Parameter Measurement

Folder: 2_Parameter_measurement

Folder: 2_Parameter_measurement

This folder contains code and data for measuring ice thickness, electron dose, and CTF parameters (defocus and experimental B-factor) from the cryo-ET tilt-series of  experimentally acquired POPC liposome pairs.

Subfolder: 2_1_Ice_thickness_electron_dose_measureExp_avg_ice_Pair1_region_1_2.mat: Average intensities of the 1st and 2nd ice region images from the Pair 1 tilt-seriesAngles_ice_Pair1_region_1_2.mat: Tilt angles of the 1st and 2nd ice region images from the Pair 1 tilt-seriesExp_avg_ice_Pair1_region_3_4.mat: Average intensities of the 3rd and 4th ice region images from the Pair 1 tilt-seriesAngles_ice_Pair1_region_3_4.mat: Tilt angles of the 3rd and 4th ice region images from the Pair 1 tilt-seriesExp_avg_ice_Pair2.mat: Average intensities of ice region images from the Pair 2 tilt-seriesAngles_ice_Pair2.mat: Tilt angles of ice region images from the Pair 2 tilt-series.run 'Ice_thickness_dose_measurement.m' to measure ice thickness, the tilt angle of the ice slab, and electron dose based on the average intensities of ice region images.

Subfolder: 2_2_CTF_estimationSubfolder: Carbon_images_cutofflow_Pair1: Carbon region images from the Pair 1 tilt-series and the lower bound of spatial frequency for power spectrum fitting.Subfolder: Carbon_images_cutofflow_Pair2: Carbon region images from the Pair 2 tilt-series and the lower bound of spatial frequency for power spectrum fitting.Carbonimages_positions_tiltangle_Pair1.mat: Coordinates of carbon region images from the Pair 1 tilt-series and the corresponding tilt angles.Carbonimages_positions_tiltangle_Pair2.mat: Coordinates of carbon region images from the Pair 2 tilt-series and the corresponding tilt angles.Liposomepair_positions_tiltangle_Pair1.mat: Coordinates of liposome pair from the Pair 1 tilt-series and the corresponding tilt angles.Liposomepair_positions_tiltangle_Pair2.mat: Coordinates of liposome pair from the Pair 2 tilt-series and the corresponding tilt angles.Carbonimage_frac_ampli_Pair1.mat: Fraction of amplitude contrast in the carbon region of the Pair 1 tilt-seriesCarbonimage_frac_ampli_Pair2.mat: Fraction of amplitude contrast in the carbon region of the Pair 2 tilt-seriesrun 'Power_spectrum_fitting.m' to compute the power spectrum from the carbon region image and estimate defocus and the experimental B-factor by fitting.

3. Simulated Cryo-ET Tilt-Series Generation

Folder: 3_Simulated_cryo-ET_tiltseries_generation

This folder contains code and data for generating cryo-ET tilt-series of POPC liposomes using TEM simulation.

Subfolder: 3_1_Atomic_structure_generation

Lownoise_POPCliposome1.mat, Lownoise_POPCliposome2.mat, Highnoise_POPCliposome1.mat, Highnoise_POPCliposome2.mat, Highnoise_POPCliposome3.mat: Atomic structure files of spherical POPC liposomes included in each liposome pair, generated using the CHARMM-GUI Martini Maker software.run 'Atomic_structure_generation_tiltseries_lownoise_Pair1.m' to generate atomic structures of low noise POPC liposome pair (Pair 1) embedded in ice, rotated at various angles.POPCliposomepair_Lownoise_atomic_structure_0degree_xyzfile.xyz: Final atomic structure file of a low noise POPC liposome pair. This folder provides only the structure at 0 degree as a representative example.Atomic structures at other tilt angles are identical except for the rotation angle.run 'Atomic_structure_generation_tiltseries_highnoise_Pair2.m' to generate atomic structures of high noise POPC liposome pair (Pair 2) embedded in ice, rotated at various angles.POPCliposomepair_Highnoise_atomic_structure_5degree_xyzfile.xyz: Final atomic structure file of a low noise POPC liposome pair. This folder provides only the structure at 5 degree as a representative example.Atomic structures at other tilt angles are identical except for the rotation angle.Deformed_POPCliposome1.mat, Deformed_POPCliposome2.mat: Atomic structure files of spherical POPC liposomes that make up the deformed liposomes, generated using the CHARMM-GUI Martini Maker software.run 'Atomic_structure_generation_tiltseries_deformed_liposome.m' to generate atomic structures of POPC deformed liposome embedded in ice, rotated at various angles.POPCliposomedeformed_atomic_structure_0degree_xyzfile.xyz: Final atomic structure file of a deformed POPC liposome. This folder provides only the structure at 0 degree as a representative example.Atomic structures at other tilt angles are identical except for the rotation angle.

Subfolder: 3_2_Multislice_simulation

run 'Multislice_simulation_TEMSIM_from_atomic_structure.sh' to calculate 2D exit wave arrays from the atomic structures.run 'Convert_Exitwavedata_to_matfile.m' to convert 2D exit wave array data into .mat file format. The output .mat file is used for TEM image formation.Exitwave_2Darray_LowNoise_Pair1.mat: 2D exit wave array of low noise liposome pair in .mat format and corresponding tilt angleExitwave_2Darray_HighNoise_Pair2.mat: 2D exit wave array of high noise liposome pair in .mat format and corresponding tilt angleExitwave_2Darray_Deformedliposome.mat: 2D exit wave array of deformed liposome in .mat format and corresponding tilt angle

Subfolder: 3_3_TEM_image_formation

CTFparameters_Pair1.mat: Defocus and experimental B-factor of Pair 1 tilt-series.CTFparameters_Pair2.mat: Defocus and experimental B-factor of Pair 2 tilt-series.run 'CryoET_tiltseries_generation_from_exitwave.m' to generate tilt-series of POPC liposomes from the 2D exit wave arrays.Simulated_cryoETtiltseries_LowNoise.mat: Simulated noisy and ground truth cryo-ET tilt-series of a low noise liposome pairSimulated_cryoETtiltseries_HighNoise.mat: Simulated noisy and ground truth cryo-ET tilt-series of a high noise liposome pairSimulated_cryoETtiltseries_Deformedliposome.mat: Simulated noisy and ground truth cryo-ET tilt-series of a deformed liposome

4. Training Set Generation

Folder: 4_Traning_set_generation_via_TEMsimulation

This folder contains code and data for generating a training set of POPC liposomes using TEM simulation.

Subfolder: 4_1_Atomic_structure_generation

Lownoise_POPCliposome1.mat, Lownoise_POPCliposome2.mat, Highnoise_POPCliposome1.mat, Highnoise_POPCliposome2.mat, Highnoise_POPCliposome3.mat: Atomic structure files of spherical POPC liposomes included in each liposome pair, generated using the CHARMM-GUI Martini Maker software.Deformed_POPCliposome1.mat, Deformed_POPCliposome2.mat: Atomic structure files of spherical POPC liposomes that make up the deformed liposomes, generated using the CHARMM-GUI Martini Maker software.run 'Atomic_structure_generation_traningset_Lownoise_Pair1.m' to generate atomic structures of various-sized low noise POPC liposome pair (Pair 1) embedded in ice and rotated at various angles.POPCliposomepair_Lownoise_atomic_structure_sample1_xyzfile.xyz: Final atomic structure file of a low noise POPC liposome pair. This folder provides only the 1st structure as a representative example.run 'Atomic_structure_generation_traningset_Highnoise_Pair2.m' to generate atomic structures of various-sized high noise POPC liposome pair (Pair 2) embedded in ice and rotated at various angles.run 'Atomic_structure_generation_traningset_deformed.m' to generate atomic structures of various-sized deformed POPC liposome embedded in ice and rotated at various angles.

Subfolder: 4_2_Multislice_simulation

run 'Multislice_simulation_TEMSIM_from_atomic_structure_trainingset.sh' to calculate 2D exit wave arrays from the atomic structures.run 'Convert_Exitwavedata_train_to_matfile.m' to convert 2D exit wave array data into .mat file format. The output .mat file is used for TEM image formation.Exitwave_2Darray_train_LowNoise_Pair1.mat: 2D exit wave array of low noise liposome pair in .mat formatExitwave_2Darray_train_HighNoise_Pair2.mat: 2D exit wave array of high noise liposome pair in .mat formatExitwave_2Darray_train_Deformedliposome.mat: 2D exit wave array of deformed liposome in .mat format

Subfolder: 4_3_TEM_image_formation

run 'CryoET_traningset_generation_from_exitwave.m' to generate training sets of POPC liposome cryo-EM images from the 2D exit wave arrays.

5. U-net CNN Training and Image Denoising

Folder: 5_Unet_CNN_training_image_denoising

This folder contains code and data for training a convolutional neural network (CNN) with a U-net architecture using a training set of POPC liposome cryo-EM images, and for denoising images using the trained CNN.

run 'Unet_CNN_training_from_cryoEM_images.py' to train a U-net CNN using a training set of POPC liposome cryo-EM images.

Subfolder: 5_Results_trainedCNN_LowNoise_Pair1Subfolder: 5_Results_trainedCNN_HighNoise_Pair2Subfolder: 5_Results_trainedCNN_Deformedliposome

run 'Plot_loss_curve.ipynb' to plot the training loss as a function of epoch.run 'Image_denoising_using_trainedCNN.ipynb' to denoise noisy cryo-EM images using the trained CNN.Denoised_images_LowNoise.mat: Denoised simulated cryo-ET tilt-series of low noise liposome pair.Denoised_images_HighNoise.mat: Denoised simulated cryo-ET tilt-series of high noise liposome pair.Denoised_images_Deformedliposome.mat: Denoised simulated cryo-ET tilt-series of deformed liposome.Denoised_images_Pair1.mat: Denoised experimental cryo-ET tilt-series of Pair 1.Denoised_images_Pair2.mat: Denoised experimental cryo-ET tilt-series of Pair 2.

6. CTF Correction, Tilt-Series Alignment, and 3D Reconstruction

Folder: 6_CTFcorr_alignment_3D_reconstruction

This folder contains code and data for performing CTF correction, tilt-series alignment, and 3D reconstruction using tilt-series obtained from simulations or experiments.

Defocus_Pair1_low_noise_pair.mat: Defocus values of the tilt-series of the low noise liposome pair and Pair 1, used for CTF correctionDefocus_Pair2_high_noise_pair.mat: Defocus values of the tilt-series of the high noise liposome pair and Pair 2, used for CTF correctionDefocus_deformed_liposome.mat: Defocus values of the tilt-series from the deformed liposome, used for CTF correctionFracampli_Pair1_low_noise_pair.mat: Fraction of amplitude contrast of the tilt-series of the low noise liposome pair and Pair 1, used for CTF correctionFracampli_Pair2_high_noise_pair.mat: Fraction of amplitude contrast of the tilt-series of the high noise liposome pair and Pair 2, used for CTF correctionFracampli_deformed_liposome.mat: Fraction of amplitude contrast of the tilt-series from the deformed liposome, used for CTF correctionMask_low_noise_pair.mat: Mask for the simulated tilt-series of the low noise liposome pair.Mask_high_noise_pair.mat: Mask for the simulated tilt-series of the high noise liposome pair.Mask_deformed_liposome.mat: Mask for the simulated tilt-series of the deformed liposome.Mask_Pair1.mat: Manually defined mask for the liposome pair boundary in the Pair 1 tilt-seriesMask_Pair2.mat: Manually defined mask for the liposome pair boundary in the Pair 2 tilt-series

run 'CTFcorr_alignment_3Dreconstruction_once_simul.m' to perform CTF correction, tilt-series alignment, and 3D reconstruction on the simulated tilt-series.run 'CTFcorr_alignment_3Dreconstruction_once_expe.m' to perform CTF correction, tilt-series alignment, and 3D reconstruction on the experimental tilt-series.

Subfolder: Reconstruction_results_low_noise_pair, Reconstruction_results_high_noise_pair, Reconstruction_results_Pair1, Reconstruction_results_Pair2, Reconstruction_results_deformed_liposome

7. 3D Reconstruction Results and SNR Calculation Plots

Folder: 7_Reconstruction_results_SNR_plot

This folder contains code and data for plotting the results of 3D reconstruction and the SNR of images and lipid bilayers in tomograms.

lownoise_tomogram_lipidbilayer_mask.mat: Outer lipid bilayer mask of the low-noise liposome pairlownoise_tomogram_interior_mask.mat: Interior region mask of the low-noise liposome pairhighnoise_tomogram_lipidbilayer_mask.mat: 3D mask of the inner lipid bilayer in the high-noise liposome pairhighnoise_tomogram_interbilayer_mask.mat: 3D mask of the inter-bilayer region in the high-noise liposome pairPair1_tomogram_lipidbilayer_mask.mat: 3D mask of the outer lipid bilayer in Pair 1Pair1_tomogram_interior_mask.mat: 3D mask of the interior region in Pair 1Pair2_tomogram_lipidbilayer_mask.mat: 3D mask of the inner lipid bilayer in Pair 2Pair2_tomogram_interbilayer_mask.mat: 3D mask of the inter-bilayer region in Pair 2deformed_tomogram_lipidbilayer_mask.mat: 3D mask of the outer lipid bilayer in deformed liposomedeformed_tomogram_interior_mask.mat: 3D mask of the interior region in deformed liposome

run 'Plot_reconstruction_and_SNR_results_lownoise_highnoise.m' to visualize 3D reconstruction results, image SNR, and lipid bilayer SNR in tomograms generated from simulated liposome pair tilt-series.run 'Plot_reconstruction_and_SNR_results_Pair1_Pair2.m' to visualize 3D reconstruction results, image SNR, and lipid bilayer SNR in tomograms generated from Pair 1 and Pair 2 tilt-series.run 'Plot_reconstruction_and_SNR_results_deformed.m' to visualize 3D reconstruction results, image SNR, and lipid bilayer SNR in tomograms generated from simulated deformed liposome tilt-series.

Additional Notes

If you use any of the above data or source codes in your publications and/or presentations, our paper should be properly cited: Y. Kim and Y. Yang, "Supervised cryo-EM tomography denoising enabled by physics-based image simulation", Appl. Phys. Lett. 128 (2026).

If you have any questions regarding the above data or source codes, please contact Yongsoo Yang, Department of Physics, Korea Advanced Institute of Science and Technology (KAIST), Daejeon 34141, Korea. Email: [e-mail] . The code is distributed under the terms of the GNU General Public License v3.0 (GPL-3.0).