What shapes template-matching performance in cryogenic electron tomography in situ?

Area of Science:

• Structural biology
• Biophysics
• Computational biology

Background:

• 3D template matching is crucial for detecting biological macromolecules in cryo-electron tomography (cryo-ET) data.
• High binning is commonly employed to reduce computational costs and noise during template matching, but its impact on accuracy is not fully understood.

Purpose of the Study:

• To systematically evaluate the influence of template size, shape, and angular sampling on macromolecule detection using 3D template matching in cryo-ET.
• To identify limitations of current template matching practices, particularly at high binning levels.

Main Methods:

• A ground-truth annotated dataset was used to analyze the relationship between template parameters and detection performance.
• Theoretical considerations were applied to explain the observed experimental results.

Main Results:

• At commonly used high binning levels, detailed subtomogram averages, spheres, and even a heart emoji exhibited nearly identical performance in template matching.
• This indicates that current methods struggle to precisely detect macromolecules unless their shape and size are significantly distinct from the background.

Conclusions:

• High binning in cryo-ET template matching predominantly retains low-frequency information, leading to similar spectral representations for similarly shaped and sized macromolecules.
• This inherent limitation hinders accurate detection and necessitates the development of enhanced template matching methodologies for improved precision in structural biology.