CryoPROS: Correcting misalignment caused by preferred orientation using AI-generated auxiliary particles.

Area of Science:

• Structural Biology
• Biophysics
• Computational Biology

Background:

• Preferred orientation is a common challenge in cryo-electron microscopy (cryo-EM) that hinders accurate 3D structure reconstruction.
• Conventional methods struggle to overcome the data anisotropy caused by preferred particle orientation.

Purpose of the Study:

• To introduce cryoPROS, a novel computational framework for correcting preferred orientation in cryo-EM data.
• To enhance particle alignment accuracy using self-supervised deep generative models for auxiliary particle generation.

Main Methods:

• CryoPROS co-refines raw and generated auxiliary particles to correct misalignment.
• Auxiliary particles are created using a self-supervised deep generative model.
• The framework was tested on various datasets, including HA-trimer, P001-Y, NaX, and hormone-sensitive lipase dimer.

Main Results:

• CryoPROS achieved near-atomic resolution on an untilted HA-trimer dataset.
• High-resolution structures were successfully resolved from three experimental datasets affected by preferred orientation.
• Extensive validation confirmed the robustness of cryoPROS and its low risk of introducing model bias.

Conclusions:

• Addressing misalignment due to preferred orientation can significantly improve cryo-EM density maps.
• CryoPROS offers a robust solution for datasets previously considered problematic due to preferred orientation.
• The findings suggest that preferred orientation may be overcome in many cryo-EM studies with appropriate computational strategies.