A self-supervised workflow for particle picking in cryo-EM.

Area of Science:

• Structural Biology
• Biophysics
• Computational Biology

Background:

• Accurate particle picking is crucial for high-resolution single-particle cryo-electron microscopy (cryo-EM) data analysis.
• Current methods often require significant user intervention, hindering high-throughput analysis.

Purpose of the Study:

• To develop and validate a self-supervised workflow for automated particle picking in cryo-EM.
• To reduce user input and improve the efficiency of particle selection for high-resolution structure determination.

Main Methods:

• Implemented a self-supervised workflow featuring an iterative strategy using 2D class averages to refine training particles.
• Employed a progressively improved convolutional neural network (CNN) for particle identification.
• Defined an automated particle selection threshold using the ratio of percentage class distribution and resolution (%/Res).

Main Results:

• Successfully tested the workflow on six diverse, publicly available cryo-EM datasets.
• Demonstrated automatic particle picking with minimal user intervention across various particle sizes and shapes.
• Achieved particle selection that supports high-resolution cryo-EM reconstructions at 3.0 Å or better.

Conclusions:

• The developed self-supervised workflow represents a significant advancement towards automated particle picking in single-particle cryo-EM analysis.
• This method can be integrated with existing cryo-EM software packages like Relion, cryoSPARC, cisTEM, SPHIRE, and EMAN2.
• The workflow facilitates efficient and accurate particle selection, enabling high-resolution structural biology studies.