3dem
Symmetry file input is nonfunctional
RELION retains the functionality of XMIPP to provide a symmetry definition file as input to the --sym flag, as described here.
However the healpix-tesselation is reduced based on the applied symmetry, which is limited to the enumerated point-groups and their conventional orientations as defined in RELION. Hence, any custom input file dies with
ERROR: Symmetry -1is not known
thrown from src/healpix.cpp
It would be nice if the logic behind the healpix-reduction dependent on the applied symmetry axes was generalized so that custom alignment symmetries were permissible. Said logic would arguably be complicated, but at the moment there is no correct manner in which to e.g. use the --relax_symmetry option to sub-symmetries for which relion does not have aligned definitions. If One tries to relax icosahedral to tetrahedral or D2, then the reference alignment, particle alignment parameters and healpix reduction are all incompatible.
It's tempting to say that custom symmetries is an exotic luxury, but it's a really nice way for people to control which symmetries to relax, and which sub-symmetries to investigate. Icosahedral reconstructions might have interesting features in T, D2, C5, C3 or C2, not to mention mismatching symmetries along any of those. Instead of defining all these consistently in RELION, symmetry file input is really convenient.
This is an interesting suggestion but indeed sounds like "an exotic luxury" for me to invest time. If you can implement this yourself, I am happy to review the code.
For a longer term, we are currently designing a full rewrite of relion_refine. One topic of discussion is to include a very general model of local symmetry and multibody refinement. For example, it will allow a complex of AA'BB', each of four monomers having their own poses (but restrained to each other), and A and A', B and B' sharing the same reference.
