crYOLO: Automated particle picking

crYOLO uses a pre-trained deep learning neural network to pick particles from cryo-EM micrographs.

Benefits of using crYOLO for particle picking:

  • In our hands, crYOLO works routinely on 95% of samples that we have tested without any parameter tuning.
  • Does not require user-defined diameters
  • Outputs particle diameter histogram based on the particles it ‘sees’ in the micrographs
  • Fast! crYOLO can pick a few million particles in 30-60 minutes.

Running crYOLO on COSMIC²:

The general model for particle picking works very well and we use the latest general model available from the crYOLO team. NOTE: We are using the general model for cryo-EM micrographs. Do not use denoised cryo-EM micrographs or negative stain micrographs.

Uploading micrographs to COSMIC²:

NOTE: If you have your micrographs in a RELION job directory (e.g. MotionCorr/job001/) then you can just select the directory “job001” for upload into COSMIC².

COSMIC2 uses STAR files as the data object, so for data upload we expect a STAR file in a directory containing micrographs.

  1. Create a new directory (e.g. micrographs_for_cosmic2)
  2. Place all micrographs to be classified into the new directory (e.g. ‘micrographs_for_cosmic2/‘)
  3. Create a .star file with micrograph names within micrographs_for_cosmic2 by using this command: 
    • printf 'data_\nloop_\n_rlnMicrographName\n' >> micrographs_for_cosmic2/micrographs.star | ls micrographs_for_cosmic2/*.mrc >> micrographs_for_cosmic2/micrographs.star
  4. Upload this directory to COSMIC² using Globus.
    • Note: in order to upload this directory, you will need to select the directory named micrographs_for_cosmic2 in the above example
  5. If successful, you will see a new data file within your data page
    • For this example, it will be named micrographs_for_cosmic2/micrographs.star

Job parameters

After selecting your micrographs STAR file, the only parameter to change is the threshold. The threshold value sets how stringent (or not) to select particles from micrographs. We typically do not need to change this parameter, a threshold of 0.1 works in 90% of cases.

Outputs from COSMIC²

When finished, you will see the following files listed in the output page for your task:

crYOLO coordinate output files
  • This is a .zip file containing all outputs from crYOLO. This includes particle coordinates in a few different formats (STAR, box, etc.) as well as summary information for the job (DISTR).
  • After unzipping this file you will see the following files in the unzipped folder:
    • CBOX/ – .cbox formatted particle coordinates
    • cryolo_config.json – crYOLO config file used during run
    • cryolo.err – standard error file from crYOLO
    • cryolo.out – standard out file from crYOLO. <—- This file will have summary information for your run such as particles picked per micrograph, total particles found, and average particle diameter.
    • DISTR/ – folder containing size distribution histogram and confidence distribution
    • EMAN/ – .box formatted particle coordinates
    • STAR/ – STAR formatted coordinates

Using crYOLO picks to extract particle with RELION

After running crYOLO, you can extract particles using RELION.

Protocol:

  1. Download & unzip .zip file from COSMIC²
  2. Navigate to your RELION project directory.
    • The project directory is the location from which you open the RELION GUI and also has folders named ‘Class2D’, ‘Refine3D’, etc.
  3. Copy all STAR files from crYOLO into the directory containing the micrograph or movie files that you used to import into RELION
    • E.g. Let’s say you had your micrograph files in a folder named ‘Micrographs’ and you imported these micrographs into RELION. In this case, you would copy your STAR files into this directory so that the directory has both micrograph and STAR files.
  4. Import STAR files into RELION using GUI.

License

Note that crYOLO is free for academics but requires a license for commercial use. Please read more here: https://cryolo.readthedocs.io/en/stable/

Citation

Wagner, T., Merino, F., Stabrin, M., Moriya, T., Antoni, C., Apelbaum, A., Hagel, P., Sitsel, O., Raisch, T., Prumbaum, D., Quentin, D., Roderer, D., Tacke, S., Siebolds, B., Schubert, E., Shaikh, T.R., Lill, P., Gatsogiannis, C., Raunser, S., 2019. SPHIRE-crYOLO is a fast and accurate fully automated particle picker for cryo-EM. Commun Biol 2. https://doi.org/10.1038/s42003-019-0437-z