Manual Classification

For single particle analysis, it can be valuable to sort particles within a localization dataset. Within the multicolorSPR workflow, we use a simple three-class classification for two main purposes:

1. To determine the axial translation between two imaged proteins

• For protein pairs with a shared symmetry axis, selected orientations can be used to determine the translation parameters between the respective 3D volumes. The centriole, which we have used as a model structure to develop multicolor SPR, has a characteristic 9-fold radial symmetry, i.e many centriolar proteins form rings along the longitudinal axis. In order to determine the relative position between two proteins it is thus sufficient to look only at centriolar side view projections. To isolate only single orientations we provide a simple manual classifier to group particle according to their orientation.

2. To characterize size and symmetry properties of the imaged structures.

• 2D particle averaging is a powerful tool to increase the SNR of a labelled structure as many of those suffer from heterogeneous labelling, a common caveat in fluorescence microscopy. Please see our recent review and publications mentioned therein for a broader introduction into 2D particle averaging.

Notably: While the manually classified particles can be used for any further processing step, we use the classified data subset as a so-called response to train machine classifier that are then able to perform the particle classification autonomously.

Select Particles > Manual Classification from the SPARTAN menu.

1. Load a particle dataset. The number of particles will be displayed together with the first particle.

This module offers the classification into three classes. During the multicolorSPR workflow of centriolar structures, we classify particles into top view, side view and unclassified. These are required for subsequent volume translation,