Good afternoon,
I have been having issues with the last 3 versions of Phenix (1.14, 1.15 and 1.16) when the XYZ(real-space) option is selected, version 1.13 was the last one that doesn't have this problem for me. This seems to place too much weight on correcting rotamer 'outliers' at the expense of fitting to the density. The proteins that I work on have always had some residues that consistently are identified as rotamer outliers, but these latest versions 'correct' just about all of them (typically from 1% down to 0.1%). The resulting structures show clear movement of the sidechains out of what should be their density. Image 1 (after using 1.13) and 2 (after using 1.14, 1.15 or 1.16) show a really bad example where the Arg sidechain even gets broken apart to be moved into the stronger density of a neighbouring symmetry Gln. Image 3 (after using 1.13) and 4 (after using 1.14, 1.15 or 1.16) is a more subtle example, but although the residue is no longer a rotamer outlier after refinement using 1.16, it clearly does not fit the density as well as after using 1.13. These subtle changes are particularly worrying, as they could easily be missed when checking the structure after refinement. Other members of the group have also noticed this problem for different proteins.
So currently I have a choice of either using version 1.13 (which with Onedep no longer accepting pdb files makes things a bit inconvenient as is nice to use the latest version for getting the mmcif and validation), or to use 1.16, but switch off the XYZ(real-space).
Has anyone else noticed this happening? Found a solution to the problem?
Thank you for any suggestions,
Gyles Cozier.
