Hi @shashi!
This is a good question — it might be surprising, but this is the intended behavior of Ab initio reconstruction. To understand why, consider the goals of Ab initio:
- classify particles based on which 3D map they belong to, and
- produce “good enough” starting volumes you can use in downstream jobs like Homogeneous Refinement or Non-Uniform Refinement.
If you request only a single class, goal (1) does not apply — you’re implicitly saying that all particles belong to the same volume. Thus, Ab initio only uses a subset of the particle images, since in most cases a “good enough” starting volume can be produced without using all of your particle images.
If you request multiple classes in Ab initio (which you can do by setting Number of classes to a value greater than 1), then all particles will be used by default in order to accomplish goal number 1.
So, typically, if your particles likely contain junk picks, you’d set Number of classes to a larger value (say, 3 or 4). All of the particles would be used, and you’d keep only the classes which look how you expect for your target.
If, on the other hand, your particles are thought to be mostly clean, you’d typically set Number of classes to 1. This would not use all your particles, but you’d then run a Homogeneous Refinement using all of your particles and the Ab initio volume. This gives you a better result than running Ab initio with all of your particles would have, and it makes the Ab initio job run much faster.
I hope that is helpful!