Hi,
When analyzing complex mixtures including species with multiple different symmetries, it would be useful to be able to enforce symmetry on a per-class basis. E.g. to allow more effective multi-reference refinement of multiple unrelated species, or to allow separation of symmetric & asymmetric/pseudosymmetric versions of the same entity; or to allow decoy classification where symmetry is enforced for the good class, but not for the random density decoys.
Cheers
Oli
yes, I second Oli and would also advocate for this desirable function 
Best, Matthias
Hi @olibclarke!
Just to be sure I understand, you’d like to be able to run a 3D Classification job with some user-defined number of classes to have some (also user-defined) non-C1 symmetry while the rest are C1?
Could you explain a bit more about your imagined workflow for this? I’m trying to understand how much we’d expect classification to improve with imposed symmetry on certain classes and wonder if you could explain your thinking (or if you’re thinking of another benefit I’m missing).
Hi Rich,
Many cases would benefit from this, I think. In terms of imposing a single point group symmetry, we often find this beneficial in classification (either heterogeneous refinement or without alignments) when we have minor populations that have the same point group symmetry as the dominant class, but different conformations - think the open & closed states of a tetrameric ion channel. Often in such cases C1 classification struggles, with the minor class with limited number of particles not being well resolved.
However, in many cases the minor class may not have the same symmetry as the major class - think C2 and C4 symmetric states of an ion channel, for example. In such cases I would expect classification to benefit from imposing the correct symmetry on the minor class.
There are other cases that would benefit too - think about molecules with variable cyclic symmetry, for example gasdermins, where one can have a mixture of for example C27 & C28 symmetric species. Also true for some pore forming toxins. I would expect classification in such a case to benefit substantially from imposing the correct (or perhaps candidate) symmetries on the respective references.
There are other examples too - for example mixtures of enzymes with different point group symmetries - but there are enough cases that I think it would be a useful tool to have available.
Another case - let’s say you have a mixture of symmetric oligomers with mild preferential orientation (and different point group symmetries). Enforcing symmetry can mitigate density distortion caused by orientation bias, but if you have a mixture of symmetries there is currently no way to do this on a per class basis - I would be surprised in such a case if enforcing appropriate symmetry per class did not improve classification results.
I would also say that mixtures of species with different point group symmetries will become more common when dealing with samples from native cellular fractions, e.g.: High-Resolution Structural Proteomics of Mitochondria Using the ‘Build and Retrieve’ Methodology - PMC
Cheers
Oli
Thank you for the detailed explanation! I had thought of the ion channel example (which, in the example you gave, I might think would work well when imposing C2 on the entire classification), but the other examples you gave would clearly require different symmetries for each class. We’ve recorded this feature request!