Learning to automate cryo-electron microscopy data collection with Ptolemy.

Area of Science:

• Structural Biology
• Biophysics
• Computational Biology

Background:

• Cryo-electron microscopy (cryoEM) is crucial for determining near-atomic 3D structures of biological macromolecules.
• Increasing demand for cryoEM necessitates automated methods to improve throughput and efficiency.
• Current cryoEM data collection software often requires time-consuming manual parameter tuning or target selection.

Purpose of the Study:

• To develop automated methods for cryo-electron microscopy (cryoEM) data collection.
• To address the challenges of low signal-to-noise ratios and variable experimental parameters in automated targeting.
• To create a versatile pipeline for efficient and automated cryoEM target selection.

Main Methods:

• Developed Ptolemy, a pipeline for automated low- and medium-magnification targeting in cryoEM.
• Utilized computer vision and machine learning algorithms, including mixture models, convolutional neural networks, and U-Nets.
• Trained models on a large dataset of real-world cryoEM images labeled by human operators.

Main Results:

• Ptolemy accurately detects and classifies regions of interest in cryoEM images.
• The pipeline demonstrates generalization to unseen data collection sessions and different microscopes.
• Achieved automation of cryoEM data collection, improving efficiency and throughput.

Conclusions:

• Ptolemy provides an open-source, modular solution for automating cryoEM data collection.
• The pipeline can be integrated with existing microscope control software.
• Ptolemy serves as a foundation for future advancements in cryoEM automation.