I combined maps from a non-uniform refinement and a local refinement (in ChimeraX, using volume multiply then volume maximum) to build my my model. Is there a best practice for preforming model validation (Phenix) on a model built from a combined map, specifically in terms of half maps? Should I be combining half-maps A/B from the NU refinement and local refinement? Or splitting my model up to “fit” each individual component map?
For those interested, this is a ribosomal map - the NU refinement got me great resolution for the body of the small subunit while the local refinement focused on the small subunit head (very mobile).
Most conservative: build only in individual component maps, deposit individual component maps into the EMDB, only use composite map for figures (and make it clear what you are using)
Also OK, in my opinion (although this varies person-to-person): use phenix Combine Focused Maps to combine your maps, and give the utility your half-maps so it will generate half-maps in return (ideally you will have the same independent half-sets that were used to create both input maps). It is a little tricky to get the GUI to output half-maps, but it will do it and then it plays nicely with Phenix, although you may have to estimate the weight in Coot before refinement because a user of mine is finding that the scaling appears to be a bit different post-combination. In this case you should still deposit the separate maps that were used to generate the combined map, but they can be as auxillary uploads to the main map.
Some folks are strongly against combining focused maps because the output map doesn’t truly reflect a experimental state and the crossover regions between focused maps can get a bit distorted, but my opinion is that as long as you are very clear what you did and give enough data that someone can reproduce it / make their own conclusions, you can combine.
5 Likes