Hi James, quick answer: no, there is no option in phenix.refine to do this. Longer answer: Work in this direction sort of stalled halfway between proving the principle and putting it into active production (refinement). We use this definition of total model structure factors: Fmodel = scale * (Fatoms + Fbulksolvent) scale is an anisotropic resolution-dependent scale factor, Fatoms = Fcalc from the atomic model, Fbulksolvent = contribution of bulk solvent modeled using a flat mask (implementation specifics: https://doi.org/10.1107/S0907444913000462). Now, clearly one may wish to have an extended version of it: Fmodel = scale * (Fatoms + Fbulksolvent + F1 + F2 + ... + FN) where examples of F1, F2, ... are listed here: https://doi.org/10.1002/pro.4909 While it looks trivial to just add these components F1, F2, ... to Fmodel, it turned out to be quite challenging to do so in a numerically stable and computationally efficient manner. We solved this challenge, implemented the relevant algorithms in CCTBX, and published the methods here: https://doi.org/10.1107/S205327332300356X Now, this stalled at the point where we have phenix.mosaic, which finds and treats isolated solvent regions independently as those F1, F2, ... but: a) there is no way for you to provide the annotation for these regions, and b) this hasn’t gone into refinement yet. As you hopefully see, the above is perhaps 95% of what’s needed to enable you to do what you want in phenix.refine. Someone just needs to do the "remaining 5%"! If you convince me (e.g., this is needed for an exciting project at hand), I might be able to make another push and implement it. All the best! Pavel On 8/29/25 13:39, James Holton wrote:
Greetings PhenixBB,
I have a question that keeps coming up again and again. If I have a part of my structure that is a blob of density and I have a map of what that density cloud looks like, is there any way in phenix to include that blob as part of the refinement? Perhaps giving it its own scale and B factor? One example might be a user-defined bulk solvent map, but a disordered ligand or domain can also come up. Traditionally, in older refinement programs this was called a "partial structure", and it is a feature still supported in refmac5. You give it "FPART=F PHIP=P" in the input mtz file.
Last I checked there was no way to do this in phenix. Is that still true? Any recommended ways to do things like this?
Thanks for your help and insight,
-James Holton MAD Scientist