Hi,
When I attempt to launch local refinement, starting from an ab initio run, I get the attached error. Can local refinement not be started from an ab initio run?
The reason I’d like to do this is that sometimes, ab initio reconstruction to high resolution gives a better map than homogeneous or non uniform refinement, so I’d like to try local refinement starting from the orientation parameters of a successful high resolution ab initio run.
Cheers
Oli
Hi @olibclarke,
Thanks for pointing this out. You are right, this is currently not possible. We will address this in a future update.
For now, I would recommend reducing the “Initial Lowpass Resolution (A)” parameter in the refinement job you are running after the ab-initio job. This may improve the result of the refinement by allowing more information to be carried forward from ab-initio.
Are there any special characteristics of these datasets that tend to give better results in ab-initio? I am wondering what the cause of this might be.
Best,
Ali
Hi Ali,
Thanks - yes I have tried that and in some cases it works, but not always. We’ve encountered this scenario a couple of times (and I believe others have too) with small membrane proteins, where we can see very good high resolution features in ab initio, but can then never get results that are as good as the ab initio from subsequent refinement (even after playing around with the lowpass res and masking parameters).
I ascribed it to the fact that ab initio uses stochastic gradient descent rather than branch-and-bound, and so was wondering if in such a case ab initio (to get the approximate orientations) followed by local refinement might work better.
Cheers
Oli
