Hi all,
Using NU/ homo refinement, the job stops when the curves are still way above the 0.143 FSC, and the job “completes” when the resolution is 5.4A. The map’s resolution itself looks much lower, say around 15-20A or so.
I had 15m particles which were used to build either 1 ab initio model, or 5 ab initio models followed by hetero refinement and then performing the NU on the one that gave the best resolution.
Has anyone had this problem? Which parameters can be changed?
Here’s an example of one of the jobs -
Thanks 
Could be other things, but duplicate particles in the stack are a common culprit when you see FSC weirdness like this - the two half-sets are no longer independent (as the same particle image is in both stacks) so the FSC doesn’t converge nicely to zero.
You can use the “Remove Duplicate Particles” job (I believe under “Utilities”) to get rid of these, and then run the refinement again.
I usually run this job before any ab-initio/3D refinement - don’t do it immediately after picking/extraction as the duplicates can be ‘introduced’ (for want of a better word) during the coordinate re-centering during 2D classification (a toggle to do this as part of the 2D classification job would be a nice feature, @team!)
Thanks…
I tried as you suggested and ran the particles from a 2D select job (the one used later for the ab initio modelling) - but the “remove duplicates” job didn’t find any duplicates.
Any other suggestions? should I use those particles anyway?
Thanks again
Hi @Amitm22,
Your particles are binned (that is why FSC is not dropping below 0.143, and the Nyquist is at 5.4 Å). I would re-extract with less binning and try again.
Having said that, you should still see some details (at least secondary structural features) in the map if all is well, even with binned particles. Did the ab initio volumes look ok?
Cheers
Oli
