Recentering 2D class averages

Hi, I’ve been playing around with the new 2D classification experiment.

I tried it with some old RyR data (see attached), where the particles had been re-extracted and reccentered after refinement in relion. I noticed that a few of the classes are substantially off center, which is unexpected given the particles were pre-centered. I’ve highlighted a few of the mis-centered classes.

Is there any way to ensure that the output class averages are recentered based on the center of mass of the density? This would be desirable if one wants to use them for particle picking, so that the picked particles are not systematically mis-centered.


1 Like

Hi @olibclarke,

Thanks for pointing this out. Re-centering based on the center-of-mass is likely a good solution. Ab-initio 3D reconstructions actually also do this by default since they suffer from the same under-constrained position of the density within the box.
Nice looking classes by the way! :slight_smile:


Hi @apunjani,

Thanks for the reply! Actually in 0.5.3 I have an ab initio reconstruction which does not recenter correctly by default, and then align symmetry doesn’t seem to be able to handle non-centered reconstructions.

I got around it by repeating ab initio in C4, but maybe it might also be worth adding a “recenter” option to the “Sharpen map” tool?

Also, maybe it would be worth renaming the “Sharpen” task to something like “Modify map” - it currently does a lot more than sharpening (e.g. flipping hand etc).