My 2D classification right before my homogeneous refinement normally has great resolution. Within 50 classes total, I usually select classes with 3-6 A that look like my protein. From there I run Ab-initio, 3D classification, then homogenous refinement. I then end up with a resolution from 8-14 A instead of that 3-6 A I started with. What can I do to increase my resolution and why does this keep happening?
-It’s better to do several times 2D classification and select, which in order to make sure your final classes containing the real good particles. -Some proteins have different conformations, which can also affect the result.
Generally the most robust pipeline in cryosparc is approximately 2d followed by ab initio with 2-5 classes and then heterogenous refinement using the ab initio output followed by non-uniform refinement of the best hetero outputs.
3d classification jobs come later; they assume fixed pose/shifts front the input and strictly class. Heterogeneous refinement optimizes poses/shifts at the same time as classification, so early on it does a better job of getting rid of junk and separating out wildly different classes.
Non-uniform refinement does a much better job than homogenous refinement if there’s any degree of flexibility/ heterogeneity (ex. detergent micelle). It is slower though.
I’d do some hetero refine and non-uniform. Depending on how that turns out/ details of your data, you might benefit from playing around with job parameters (starting, ending res etc).
Another reason for good 2d and bad 3d might be orientation bias, which would probably best be solved by either sample prep optimizations or worst case collecting tilted data