3D ab initio generate model with different hands

Dear cryoSPARC users,

I am wondering is there anyway (or need) in flipping the hand of particles?
I have a flexible particle that tends to be reconstructed in different hand from time to time.
Recently, one of the Ab initio 3D reconstruction with four models yield two classes (#0 and #2 with 183K and 186K particles, respectively) that looked similar but are flipped image of each other (which is confirmed in Chimera). Homo- and NU-refinement of volume_0 using particles from both classes gave me a resolution of 3.6Å which is only slightly better than similar refinements of volume and particles from each individual class (3.75Å and 3.85Å for class_0 and class_2, respectively). Since there is a big difference in the number of particles (360K vs 180K) and only a slight improvement in the resolution, I am wondering whether we have to flip the hand of the two class of particles or CS would automatically take care of the difference.

Thank you so much for enlightenments.

You can flip the hand of volumes (not particles) using Volume Tools - flip hand in cryosparc.

@adrian, my problem is that half of the particles are in the right hand and half in the flipped hand. I would like to use both classes of particles for the refinement of one volume in either hand. I can alway flip the hand of the volume after refinement. I read in an earlier thread that poem could flip the hand of particles. @DanielAsarnow, could you please comment on this.

I do not think is It possible for a particle to have the wrong hand.
Having right or wrong hand after ab initio reconstruction is just random.

@Ricky Just refine all the particles together with one reference.

@DanielAsarnow @Adrian. Thank you both for the comments. That is exactly I did. But then I also refine each reference with the respective class of particle. My concern was that doubling the number of particles from ~185k to 390K only improved the resolution from 3.8Å for the latter to 3.6 Å for the former. Further analysis showed that indeed there are subtle differences between those two reference volumes. That could explain the limited improvement in resolution with a big increase in particle numbers. Thanks again.

Some structures - or even certain classes - will tend to converge to a certain handedness for idiosyncratic reasons. Since you were able to run a consensus refinement, you could now restart classification at a somewhat higher resolution, perhaps even using both references aligned and in the physically correct hand. If the starting resolution is 15A, say, the hand should be stable. Using a higher initial resolution is fine if your refinement is already significantly beyond.