The Shape of Things in Cryo-ET: Why Emojis Aren’t Just for Texts

Detecting specific biological macromolecules in cryogenic electron tomography (cryo-ET) data is frequently approached by applying cross-correlation-based 3D template matching. Structural models or simple geometric shapes such as spheres, cylinders, or rectangles are frequently used templates. To reduce computational cost and noise, high binning is used to aggregate voxels prior to template matching. At high binning primarily low-frequency information remains and it has been shown that template matching fails if different macromolecules have overlapping low-frequency spectrums. Here, we combine these ideas and show in theory why geometric shapes can be used as templates and validate our findings using a detailed subtomogram average, a sphere, and the so-called poop emoji ( ) as templates to identify ribosomes in an annotated S. cerevisiae dataset. Our findings indicate that with current template-matching methods, macromolecules can only be detected with high precision if their shape and size are sufficiently different from the background, or if they are present in significantly higher abundance than other macromolecules with similar features. This implies that template matching under these conditions has by design low precision and recall. We discuss these challenges and propose potential enhancements for future template matching methodologies.