On Sat, Jul 16, 2011 at 9:00 AM, Yuri wrote:

It looks like phase_and_build only takes reflections input as F SIGF. I had to use a reflection converter as I had a .sca file with I SIGI data.

Okay, this probably means that the reflection file converter added new R-free flags - or if not, phase_and_build certainly would have. As a result, you've run a refinement with one set of flags, then rebuilding (which includes more refinement) with another set of flags. I'd recommend starting back with phase_and_build again, using the model from the first refinement, but this time with the "_data.mtz" file that came with it for your reflections. Then use those reflections with all subsequent refinements. (This isn't always necessary - if your refinement still has a way to go, one round with simulated annealing can help de-bias R-free - but it's the most conservative approach. If you decide to continue with the existing model, however, use the R-free flags that phase_and_build used - either reflections.mtz if it has them, or exptl_fobs...mtz [making sure *not* to use the "experimental" phases it contains].) -Nat