This challenge aims to motivate the cryo-EM community to carefully consider approaches to single-particle processing for datasets containing preferred orientation, especially with regard to biased estimates of state populations.

We seek to assess the accuracy of current methods for estimating state populations in the presence of a biased pose distribution. We have provided a simulated dataset of a two-state protein with a non-uniform pose distribution, including template maps. With this challenge, we hope to inspire the development of more accurate and robust methods for population estimation within the cryo-EM community. Not only is preferred orientation a common problem, but datasets rarely exhibit perfectly uniform pose distributions; most algorithms do not systematically account for this, especially when estimating populations. Lastly, we are only assessing the entanglement of pose and state populations, not reconstruction quality or resolution.

Dataset

<aside> <img src="/icons/arrow-down-line_gray.svg" alt="/icons/arrow-down-line_gray.svg" width="40px" />

A particle .mrcs stack, an associated .star file, and .mrc template maps can be downloaded via Globus here (8 GB).

</aside>

This dataset contains a single two-state protein with ‘open’ and ‘closed’ template maps provided.

• 30,000 particles, 1.2 Å/pixel, 256 pixel box size
• Poses are not provided; particles are centered
• Per-particle CTF defocus values are provided

Evaluation

Entries will be evaluated by how accurately they estimate state populations within the dataset as a whole and, if applicable, how accurately they estimate per particle states and poses (i.e., alignments).

Submission

<aside> <img src="/icons/folder_gray.svg" alt="/icons/folder_gray.svg" width="40px" />

Entries can be uploaded and submitted via Google Forms here. An example submission can also be viewed here:

example_submission.pdf

</aside>

Please provide:

• Workflow summary

• ‘Open’ and ‘closed’ state population fractions

• Metadata (.star or .cs file) containing the ‘open’ state particle subset with poses and unmodified particle image paths

• Metadata (.star or .cs file) containing the ‘closed’ state particle subset with poses and unmodified particle image paths

  <aside> <img src="/icons/warning_gray.svg" alt="/icons/warning_gray.svg" width="40px" />

  All particles must be classified into an ‘open’ or ‘closed’ state.

  Optionally, particle labels may be included as ‘soft’ assignments in the range [0, 1], with 0.0 being ‘closed’ and 1.0 being ‘open’. These labels can be specified in a column of the submitted metadata files; please note the specific column name in your submission.

  Do not restack or rename particle stacks. Image paths from the provided .star file will be used to evaluate particle class assignments.

  The following command may be helpful:

  relion_align_symmetry --i *input* --o *output* --sym C11
  

  </aside>

See the submission form for more details.

Questions

For most general questions, please use the challenge discussions space. For urgent or private questions regarding this dataset, please contact Pilar Cossio ([email protected]) or Sonya Hanson ([email protected]) at the Flatiron Institute.