Hello PhenixBB I am dealing with a structure that is twinned and would like to ensure that the Rfree set I'm using is unbiased. This is what I have done so far. - Generated an initial model after Phaser MR using Autobuild. The input "Experimental Data" for autobuild was an XDS_ASCII.hkl file from data processed with XDS (along with the phaser output mtz as the "initial map"). Autobuild generates the file "exptl_fobs_phases_freeR_flags.mtz" containing the Rfree set, which I have used for all subsequent rounds of Phenix.refine. - I have built the model to completion with Coot and Phenix.refine, without applying the twin law. Only at the final round of refinement did I apply twin refinement, and it is obvious that the data is twinned (Xtriage also suggests twinning). My question - when the Rfree is established during Autobuild, does Phenix take into account all of the possible twin laws, so that the free and working sets are not related in any way by the possible twin laws? I believe Phenix does do this, but I would like to make sure. Space group is P 1 21 1 and the twin law is pseudo-merohedral h,-k,-h-l, in case those are relevant. Thank you. Matthew Bick