Ab-initio performance based on box size

Hello everyone,
Working on my last project, I realised that when I use a box size that is much larger than my particle size (270A particle size, 900A box size down-sampled to 450A), Ab-initio nicely allocates a large majority of my already clean particles (>80%) into one class and the rest into the remaining two. When I tried to do another Ab-initio with a box size that is optimal to the particle size based on the commonly used formula (450A), it allocates an equal amount of particles into each of the three classes.

I’ve reproduced this with various box sizes, and have come to the conclusion that a large box size in beneficial in better separating particles into classes during Ab-initio, rather than using an optimal box size. Does anyone have an explanation why this might be?

Many thanks in advance

Hi @Ufuk,

Thanks for your post. This is a very interesting observation and I have a couple follow-up questions.

  1. What were you attempting to do when you found this?
  2. Is this something that you have replicated across other datasets?
  3. Are the micrographs sparse or very crowded (in terms of number of particles)?