RBMC: How does it handle symmetry expanded particle sets?

Hi CryoSPARC team,

Quick question; it’ll be a couple of days before I can try myself, but am curious… how does RBMC motion estimation cope with symmetry expanded data? Does it ignore particles which overlap for purposes of motion hyperparameter estimation, or should I worry about bias due to using the same particles multiple times? (Actually, that would be interesting… would the same particle calculated twice give statistically different hyperparameters… something to test in the future…) Would a way around that be increasing the number of particles used for hyperparameter estimation?

A complex I’m trying to neaten up is (pseudo)-C2 symmetry, behaves nicely as a whole and gets to higher resolution than I expected, but some of the outlying domains are rather flexible. Having said that, they clarify quite nicely after symmetry expansion and local refinement, but I want to see if RBMC would just polish (pun not intended!) them up a little better. I’ll likely be re-centring, re-extracting and doing focussed refinements on each component of the complex, as box size is prohibitive for running RBMC on the full complex. :frowning:

Thanks for any insights. :slight_smile:

Answer: badly.

Symmetry expanded dataset crashes with the following error:

[CPU: 219.2 MB Avail: 329.25 GB] DIE: particle pairwise distance matrix is singular, check for duplicate particles (sgetrf failed)

Not unexpected, as this was the core region and thus, no dislocation/shift for focussed/local refinement/re-centering.

Will test with extremity of the complex as well shortly.

edit: Decided to edit rather than push this back to the top for no real reason, but wanted to note…

Symmetry expanded extremity RBMC crashes with the same error. Oh well, such is.

edit 2: and it got pushed to the top again anyway…? :thinking:

Hi @rbs_sci, as you’ve now discovered the hard way (:laughing:) RBMC is going to struggle with coincident particles. A few users have pointed out that this could be a useful thing to support for a variety of reasons, and I’ve recorded it as a feature request. But as always with feature requests, I can’t promise a specific timeline.

1 Like