The 3DVA result of the 1st variability component shows that the density of one subunit decreases progressively but does not completely disappear (in a simple mode). I am wondering how we should interpret this result. Is the subunit flexible, or is it not present in some particles?
In simple mode, the (scaled) component is “simply” added or subtracted from the mean structure only. If you would change the range of the scaling factor, you would (necessarily) see the subunit disappear completely. Simple mode is really only useful for seeing what each component “means” or “is doing” - for example, your component would seem to be mostly determined by the presence or absence, or relative ordering, of this subunit.
In order to understand empirically what the actual particle subsets are like, I recommend using the “clusters” or “intermediates” modes which will actually reconstruct new maps based on subsets of the particles. In the former case, via a Gaussian mixture (fancy K-means) model in the whole latent space, and in the latter by sliding a threshold window along each component.