Does optimizing CTF params and defocus in NU-refinement jobs alter the values stored in output particles?

3D Refinement
1

Hello All,

I have a question regarding turning on the optimize per-particle defocus and the optimize per-group CTF params during an NU-refinement job in light of downstream processing.

I ran an NU-ref job with these options turned on (e.g., J1), and the resolution improved nicely. I took the output particles and ran them through a 3D Classification job (e.g., J2). I now want to take a couple of classes from this 3D Classification job that seem to have the features I need, and them through NU-refinement (e.g., J3). The question is: Do I turn on the optimize per-particle defocus and the optimize per-group CTF params options for this second NU-ref J3, or are the optimized values stored in the particle files that I’ve taken from J1?

I worry that if I unduly turn on these options in J3 I would risk overfitting the data or otherwise introduce problems. I’m not sure if J1 changed the actual CTF and defocus values that are in the particle files into the optimized ones or not.

Thank you all in advance.

Kind regards,

2

Optimised parameters are stored in a jobs output to be propagated to any downstream jobs, but optimised parameters may not be 100% optimal, if you see what I mean. Generally, if parameters are optimal (or very close) you should see, e.g. per particle defocus barely change (be a tall, sharp spike around 0 on the histogram).

What you need to be careful about is overdoing beam tilt refinements - particularly third and fourth order refinements, as they are harder to fit and can sometimes drift off… if this manifests, it’s often fairly obvious; a complete reconstruction collapse, or at the minimum a significant loss in resolution.