Classifying Particles

Particle classification involves grouping images that are similar, and separating images that are distinct. In practical use, this means that experimental projections that have the same orientation (shape) are placed within the same category for later averaging. In this case, orientation means that the particles are showing the same face to the viewer and the only difference between them is that they can be rotated by some angle in the plane of the image. The experimental projections might also be shifted relative to each other, but the centering of the experimental projections is often done before classification.

Since protein particles generally do not lie in one preferred orientation on the carbon film, many views of the protein will generally be visible. Each of these classes of experimental projections has to be defined. Furthermore, alignment of particles then has to be done for each class.

For example, a book dropped randomly on the floor will preferentially land on its front or back, and then with less probability on one of the narrow sides, and finally with even less probability on just an edge or corner. The two principal classes (front showing and back showing) will make up the bulk of the data, and, as will be shown later, by having two orientations to work with, will help create a better reconstruction than if only one view was available. Within a particular view, it is important that as many projections are used that cover the 360 rotation on the plane of the micrograph. This is particularly important when using the random conical tilt method. When the tilted projections are then used to generate a first volume, the entire space of information is available. If too few particles are used or there is a preference for one orientation over another within one class, the missing information will reduce the resolution of the calculated volume.

Multivariate Statistical Analysis

Multivariate Statistical Analysis


Creating A Model