Using non-final iteration maps from NU-refinement


Sometimes it appears visually as if the quality of the map form the penultimate or earlier iterations in non-uniform refinement is better than the final map and the FSC profile is nicer. Are there any potential issues with importing this non-final map and using it as a reference in the next refinement job?


No, there’s no issue with it. However, any perceived difference is probably totally irrelevant after the initial low-pass filter before the next refinement.

By final do you mean the sharpened or filtered map? You should always inspect the raw reconstruction map as well. Sometimes, the last iteration could have a bit of overrefinement depending on where the alignment resolution limit was set in the previous iteration. You can try a local refinement with a gentle low-pass, like 6A, and a refinement limit that’s a ~2 A lower than the resolution the FSC curve said you had.

Hi, is there a way in NU refinement to specify less iterations ? I see one can only add further iterations, but not less.

Many thanks for your help !

@marino-j no unfortunately you cannot specify to use fewer interations, since the algorithm will need to continue iterations until it detects convergence.
However as mentioned in this thread, you can always use the outputs of an earlier iteration rather than the final one. For example, you can take the half-maps from a previous iteration and connect those to an “FSC Validation” job to get a new final FSC curve (with masking, noise substitution, etc).

1 Like

OK, many thanks ! I’ll try that.