3 Denoising, Feature Segmentation, Template Matching & Subtomogram Averaging

Primary cause of high noise levels in cryo-ET tomograms

Splitting total electron dose across dozens of tilt images creates an extremely low signal-to-noise ratio, especially at high spatial frequencies

Purpose of deconvolution filtering in tomogram processing

Boosting low spatial frequencies to sharpen biological contrast while dampening uninformative high-frequency noise

Mechanism of noise-to-noise deep learning denoising

Training neural networks on paired odd and even frame tomograms to preserve consistent biological signals while removing random noise

Role of CryoCare software in cryo-ET

Applying deep learning on odd and even tomogram pairs to remove background noise and produce clear cellular volumes

How Deep-DeWedge and IsoNet overcome the missing wedge problem

Training neural networks across varied subtomogram orientations to learn biological shapes and predict missing Fourier data

Limitation of traditional manual membrane segmentation

Requires time-consuming slice-by-slice manual painting with a brush in 3D visualization software

Advantage of automated membrane segmentation tools like MemBrain

Uses trained neural network models to rapidly identify and generate 3D binary masks of cellular membranes

Purpose of MemPicker and surface density projection

Flattening traced 3D membranes to project protein density maps and analyze spatial arrangements of photosynthetic complexes

Concept of template matching in cryo-electron tomography

Sliding 3D atomic reference structures like ribosome PDB models across tomograms to locate candidate proteins via cross-correlation

Interpretation of peaks in a template matching score map

High intensity correlation scores correspond to specific 3D locations and orientations matching the reference protein

Function of DeepFinder in particle picking

Using unsupervised deep learning to identify and cluster unknown macromolecular structures without requiring prior structural templates

Primary objective of subtomogram averaging

Extracting thousands of matching 3D sub-volumes, aligning them iteratively, and averaging frames to boost signal-to-noise ratio

Resolution improvement achieved through subtomogram averaging

Elevates low-resolution 30-angstrom tomogram particles to 5-to-6-angstrom structures capable of resolving secondary features like alpha-helices

Primary software suite used for subtomogram averaging and particle refinement

RELION, an industry-standard software package for 3D single particle and subtomogram processing

Purpose of back-projection in ChimeraX visualization

Placing refined high-resolution protein maps and segmented membranes back into exact 3D coordinates to reconstruct cellular context

Distinctive biological features identified in cyanobacteria tomograms

Thylakoid membranes, membrane-bound ATP synthases, cytosolic ribosomes, glycogen storage granules, and Rubisco-filled carboxysomes or pyrenoid