T20S Proteasome: Topaz Micrograph Denoising Tutorial
Topaz denoising tutorial via the Topaz wrapper available in CryoSPARC.
PreviousT20S Proteasome: Topaz Particle Picking TutorialNextJob: Topaz Train and Job: Topaz Cross Validation
Last updated
Topaz denoising tutorial via the Topaz wrapper available in CryoSPARC.
Last updated
Preprocess the T20S subset by completing steps 1-12 in the tutorial found in the Cryo-EM Data Processing in cryoSPARC: Introductory Tutorial section found here:
Ensure that the Import Movies, and CTF estimation jobs from the linked tutorial are completed as they will be required for the Topaz Denoise job. The ouputs from these two jobs will be used as inputs for the denoising job.
Select Topaz Denoise (BETA) from the Job Builder. Drag and drop the exposures_success output from the completed CTF estimation job into the micrographs input.
Use the file browser (activated by clicking the folder icon) to locate the Topaz executable path found earlier for the Path to Topaz Executable field. Instructions on how to find the Topaz executable path can be found above.
Queue the job.
The job is using the provided pretrained Topaz denoising model on the subset of 20 micrographs from the T20S tutorial.
Once the job is completed, observe the micrograph images outputted in the log. Depending on the value of the "Number of plots to show" parameter, the job will show side-by-side micrograph comparisons where the left side features the original micrograph and the right side features the denoised micrograph. This helps determine if the denoised micrographs will be in picking or other related tasks.
Select Topaz Denoise (BETA) from the Job Builder. Drag and drop the exposures_success output from the completed CTF estimation job and the imported_movies output from the completed Import Movies job into the micrographs and training_micrographs inputs respectively.
Use the file browser to locate the same Topaz executable path used in step 2. See Step 2.2 for more details.
Queue the job.
The job is training a new Topaz denoising model on the subset of 20 micrographs from the T20S tutorial. Once the training is complete, it will use the model to denoise the input micrographs. When the denoising is completed, the job will output both the denoised micrographs and the newly trained model.
As done in step 2.5, observe the shown micrograph comparisons between the original and denoised micrographs.
When training a new model, the job will also output plots of the training and validation loss. The plots for both losses should be descending overtime. If the plot for the training loss is decreasing while the plot for the validation loss is increasing, this indicates that the model has overfit and training parameters must be tuned. The simplest approach to resolving overfitting is to reduce the learning rate.
The job will output denoised micrographs that barely look denoised. That is because the training data is the exact subset of data that is being denoised. When a greater variety of training data is used to train a model, the denoising will be much more noticeable.
Select Topaz Denoise (BETA) from the Job Builder. Drag and drop the exposures_success output from the completed CTF estimation job and the topaz_denoise_model output from the completed Topaz Denoise (BETA) job from step 3 into the micrographs and denoise_model inputs respectively.
Use the file browser to locate the same Topaz executable path used in step 2. See Step 2.2 for more details.
Queue the job.
The job is used the previously trained Topaz denoising model on the same subset of 20 micrographs. This step serves to demonstrate how to use trained Topaz denoising models. When using trained models outside of this tutorial, the input micrographs should be different from those used to trained the model.
As done in step 2.5, observe the shown micrograph comparisons between the original and denoised micrographs. The output denoised micrographs should be nearly identical to those from step 3 as the model is denoising the same micrographs using the same model as the job from step 3.
Denoised micrographs mainly serve to improve particle picking using Manual Picker and deep learning pickers such as the Topaz particle picker. Denoising micrographs have no impact on preprocessing methods such as Motion Correction and CTF Estimation nor do they affect the performance of other pickers such as Blob Picker and Template Picker. However, the denoised micrographs can serve to help visualize particles in the Inspect Particle Picks job after using any of the pickers including the aforementioned Blob Picker and Template Picker.
To observe this functionality, continue the T20S tutorial until step 9 until step 9. At step 9 of the linked tutorial, pass the denoised micrographs from step 2 of this tutorial instead of the micrographs output from the Template Picker. The T20S tutorial can be found here:
