1. Denoise

Denoising can vastly improve particle picking by helping to increase signal to noise in low-dose micrographs. Different denoising algorithms exist, include deep denoisers Topaz (Bepler et al., 2020) and Janni (Wagner and Raunser., 2020) and various Gaussian and fourier space denoisers. PartiNet implements a modified heuristic Wiener filter denoiser based on the method from CryoSegNet (Gyawali et al., 2024). PartiNet's implementation introduces multiprocessing, allowing for high-throughput denoising of large datasets, as well as saving in .mrc format if you prefer to perform picking on denoised micrographs in RELION or CryoSPARC.

Parameters

Required Parameters

Parameter	Role	Example
--source	Input folder for this stage (motion-corrected .mrc files)	/data/partinet_picking/motion_corrected
--project	Dataset root where outputs and logs are written	/data/partinet_picking

Denoised micrographs are always written to <project>/denoised/. The --source directory is never modified.

Optional Parameters

Parameter	Type	Default	Description
--num_workers	int	max available CPUs	Number of CPU workers for processing
--img_format	string	png	Output format for denoised images (png, jpg, mrc)

Input Requirements

Motion-Corrected Micrographs

Your motion-corrected micrographs should ideally meet these criteria in order for the denoising to work correctly:

• Format: single-slice MRC files from RELION or CryoSPARC
• Motion correction: Total full frame motion should be less than 100 pixels
• CTF estimation: CTF fit resolution should be less than 10 Angstroms
• Convergence: Motion correction and CTF estimation should have converged appropriately

Quality Control Check

In CryoSPARC, you can verify micrograph quality using the Manually Curate Exposures job:

• Navigate to: Processing → Exposure Curation → Interactive Job: Manually Curate Exposures
• Check motion and CTF fit parameters for each micrograph
• Remove micrographs that don't meet quality criteria

Directory Structure

Your motion-corrected directory should contain:

motion_corrected/
├── micrograph_001_fractions_patch_aligned.mrc
├── micrograph_002_fractions_patch_aligned.mrc
├── micrograph_003_fractions_patch_aligned.mrc
└── ...

Setup Instructions

1. Create Project Directory

mkdir partinet_picking
cd partinet_picking
mkdir motion_corrected

2. Transfer Motion-Corrected Micrographs

From CryoSPARC:

# Using symbolic links (faster, saves space)
ln -s /path/to/cryosparc/project/JXXX/motioncorrected/*_fractions_patch_aligned.mrc motion_corrected/

# Using rsync (copies files)
rsync /path/to/cryosparc/project/JXXX/motioncorrected/*_fractions_patch_aligned.mrc motion_corrected/

From RELION:

# Link motion-corrected micrographs
ln -s /path/to/relion/project/MotionCorr/jobXXX/*.mrc motion_corrected/

3. Run Denoising

Apptainer/Singularity

apptainer exec --nv --no-home \
    -B /data oras://ghcr.io/wehi-researchcomputing/partinet:main-singularity partinet denoise \
    --source /data/partinet_picking/motion_corrected \
    --project /data/partinet_picking

Docker

docker run --gpus all -v /data:/data \
    ghcr.io/wehi-researchcomputing/partinet:main partinet denoise \
    --source /data/partinet_picking/motion_corrected \
    --project /data/partinet_picking

Local Installation

partinet denoise \
    --source /data/partinet_picking/motion_corrected \
    --project /data/partinet_picking

Output

Directory Structure

After denoising, your project directory will contain:

partinet_picking/
├── motion_corrected/
│   └── [original MRC files]
├── denoised/
│   ├── micrograph_001_fractions_patch_aligned.png
│   ├── micrograph_002_fractions_patch_aligned.png
│   ├── micrograph_003_fractions_patch_aligned.png
│   └── ...
└── partinet_denoise.log

Log File Output

The partinet_denoise.log file provides detailed processing information:

2025-07-15 17:37:03,807 - Using 48 workers out of 96 available CPUs.
2025-07-15 17:37:03,807 - Processing raw micrographs in /data/partinet_picking/motion_corrected
2025-07-15 17:37:03,807 - Saving denoised micrographs in /data/partinet_picking/denoised
2025-07-15 17:37:03,812 - Directory ready: /data/partinet_picking/denoised
2025-07-15 17:37:49,774 - Processed image micrograph_001_fractions_patch_aligned.mrc to dest. micrograph_001_fractions_patch_aligned.png
2025-07-15 17:37:49,956 - Processed image micrograph_002_fractions_patch_aligned.mrc to dest. micrograph_002_fractions_patch_aligned.png

Congratulations! You have prepared your micrographs for picking, you can now move to particle picking with Detect.

Advanced Usage

Custom CPU Configuration

Number of CPUs used by PartiNet is controlled with --num_workers. These CPUs are automatically split between tasks to optimize for CPU utilization during denoising:

• Processing CPUs: Half of available CPUs used for denoising
• I/O CPUs: Remaining CPUs reserved for file operations
• Resource efficiency: Achieves close to 100% CPU utilization

Example with 64 CPUs:

--num_workers 64

• 32 CPUs for denoising operations
• 32 CPUs for I/O operations (reading/writing micrographs)

Example in project:

partinet denoise \
    --source /data/partinet_picking/motion_corrected \
    --project /data/partinet_picking \
    --num_workers 32

Different Output Formats

By default PartiNet outputs denoised images in png format. This is the recommended format for the Detect stage. png is lossless; micrographs are normalised from 32-bit MRC to 8-bit PNG. jpg is available but not recommended due to lossy compression.

# JPEG format (smaller file size, lossy compression)
partinet denoise \
    --source /data/partinet_picking/motion_corrected \
    --project /data/partinet_picking \
    --img_format jpg

# PNG format (default, best for PartiNet pipeline, lossless compression)
partinet denoise \
    --source /data/partinet_picking/motion_corrected \
    --project /data/partinet_picking \
    --img_format png

For use with other particle pickers (RELION, CryoSPARC, Topaz, crYOLO) mrc is also available:

partinet denoise \
    --source /data/partinet_picking/motion_corrected \
    --project /data/partinet_picking \
    --img_format mrc

What's Next

• Detect

References

Bepler, T., Kelley, K., Noble, A. J., & Berger, B. (2020). Topaz-Denoise: General deep denoising models for cryoEM and cryoET. Nature Communications, 11(1), 5208. https://doi.org/10.1038/s41467-020-18952-1

Wagner, T., & Raunser, S. (2020). The evolution of SPHIRE-crYOLO particle picking and its application in automated cryo-EM processing workflows. Communications Biology, 3(1), 61. https://doi.org/10.1038/s42003-020-0790-y

Gyawali, R., Dhakal, A., Wang, L., & Cheng, J. (2024). CryoSegNet: Accurate cryo-EM protein particle picking by integrating the foundational AI image segmentation model and attention-gated U-Net. Briefings in Bioinformatics, 25(4), bbae282. https://doi.org/10.1093/bib/bbae282