Hello,

I am working with a sub-tomogram averaging dataset of a protein on membrane using Relion. I did particle picking and 3D classification at 4xbin, then continued to 3D refine at 4xbin and the resolution reached the Nyquist. Next, I unbinned my dataset to 2xbin and re-ran 3D auto-refinement. However, at this stage, I had warnings that my angular accuracy was worse than 10 degrees in 3D refine. So, as a next step, I did 3D classification with 1 class using my 2xbin particles. I kept all parameters as default except that I enabled 'Local angular search range' in the sampling tab of 3D classification. 3D classification with 1 class did very well and improved the resolution to the bin2 Nyquist. I examined the output map and it looks reasonable and proper with no odd artifacts. Then, I passed on the particle dataset form 3D classification to 3D refine and used the output map from 3D classification as input reference to 3D refine; but 3D refine failed saying it does not have enough particles in half-datasets. So I grouped the particles using default setting on the GUI Select tab from my 3D class output. I re-ran 3D refine and also enabled local search in 3D refine by giving the same value (1.8) for 'Initial angular sampling' and 'local searches from auto-sampling'. But the resolution is much worse than what I get with the 3D classification routine. I look at the output map and it looks as my molecule should but low pass filtered. I am trying to figure out why 3D auto refine procedure gives a map that is almost 5-6 angstroms worse than the 3D classification output map. Could the grouping of particles after 3D classification be a cause? Are there any additional parameters in 3D refine procedure that I can or should tweak to make the refinement go better?  Any thoughts or tips would be very helpful. 

Thanks,
Vidya

########################################################################

To unsubscribe from the CCPEM list, click the following link:
https://www.jiscmail.ac.uk/cgi-bin/webadmin?SUBED1=CCPEM&A=1