Particle Alignment

Access: CP::3D Model::Alignment (Tab)

2D Alignment Parameters

Particle images in a stack can be aligned either to the stack sum or a set of references.  The function is selected from the pull-down menu Method:

• Refr. Match : uses the template-matching algorithm to align each particle in the stack.  A reference image stack can be defined through the Ref button.  (The reference should have the same dimension as the particle images.)
• Self Align : this method first sums up the rotational average of all particles in the stack, then uses the resulting single frame for particle alignment.  Although self-align avoids the potential problem of reference bias, a convergence to a correct alignment is not guaranteed.

Other input parameters for particle alignment include:

• Image Filtering : pre-filter particle images by a Gaussian kernel to reduce noise.
• Precision : define the angular step-size in rotational search (in degrees, for Refr. Match only.  "0": translation search only; "-1": automatically set the optimal angular step-size).
• Max. Shift : the tolerance of image off-centering in the translational search (in pixels).
• Image Masking : a circular mask that defines active pixels in the particle image alignment.
• Correlation Threshold : C.C. threshold in alignment evaluation.  Particles with scores above the threshold will be highlighted as "good ones".  Note that C.C. score alone does not necessarily guarantee the quality of a selected particle.

Activating the Align button will initiate the process of particle alignment.  Upon completion, the result will be displayed at

• Total Particles : number of particles in the original stack.
• Accepted : aligned particles that score above the threshold.

The Stat. button, when activated, will plot a chart to summarize the alignment statistics (C.C. scores and in-plane translation distribution) for further analysis.  Particle alignment parameters can be imported to and/or exported from PARTICLE, and the particle alignments can be synchronized onto its original micrograph to improve centering of the particles for re-boxing.  These functions (Import, Export, Center and Clear) are accessible at the menu-bar from CP::MenuBar::Function::Particle Alignment.

Alignment Editing

In single-particle analysis, sometimes multi-fold binned particle frames are used at the beginning to achieve higher SNR in the images as well as faster processing.  Then, in the final phase of model reconstruction and refinement, particle images at the original resolution will be restored and the alignment parameters determined from the binned frames can be assigned to the images for additional processing.  The alignment editing functions and parameters are

• Mag. Factor : the scaling factor for X/Y translation.
• Amp. Contrast : re-define the amplitude contrast parameter in particle CTF.
• Alignment Source : the input of alignment parameters from binned images.
• Image ID Source : the link between class averages and raw particle images.
• Setup : to be documented.
• Merge : to be documented.
• Reduce : to be documented.

Notes

In the conventional paradigm of single-particle classification, particle images will first be aligned and then be subjected to pixel-based PCA analysis (as implemented in the ALNC method).  Hence, the outcome of a particle classification critically depends on the accuracy of particle alignment, which might contain substantial error in cryo-EM images with intrinsically low SNR.  In contrast, the CLAX classification methods of PARTICLE do not require image pre-alignment and therefore can eliminate such error and bias.  Still, particle alignment functions have been implemented in the package to be utilized at the investigator's discretion.