Top views separation during heterogeneous refinement

Hello everyone,

I’m relatively new to the world of cryo-em so I’m still learning many things. Currently, I’m working with a membrane protein (ion channel) reconstituted into the nanodisc. I get very good 2D classes with obvious structural features. I tried the ab-initio with 2 and 3 classes. Using the 3 classes started to separate the top views from the side views so I decided to continue with the 2 classes. To further sort the particles and remove junk particles I performed heterogeneous refinement using 2 ab-initio classes and two junk models. I checked the outcome of each class by running 2D classification and I realized that my top views are going to the junk models. Is there any way to prevent top views separation at this step, like tweaking any parameters or changing the junk models?

The topic might have been addressed previously but since I couldn’t find any convincing solutions, I decided to bring it up again. Thank you so much for sharing your insights.

Just do 1 class. Until you have quality model. Then sort junk later. 1 ab initio, or NU refine all particles with ab initio reference

You can also try to rescue some of the discarded particles, by pooling them, and running a heterogenous refinement using only those particles.