Hello: I have a few questions/comments with phenix.real_space_refine on Cryo-EM structures that may need your clarification. I am running version 2.2-6143, but the previous 2.1-???? gave similar experience. Thanks! 1) ADP refinement seems to be seeded randomly among runs? With exactly same inputs, while the output coordinates (x,y,z) showed always to be exactly same, there are difference in B-factors. While most of those BFs showed marginal difference, some are actually not too small. For example: ATOM 19872 CG ASP D 609 160.261 175.386 111.504 1.00 94.06 C ATOM 19873 OD1 ASP D 609 159.374 176.102 110.995 1.00 98.69 O ATOM 19874 OD2 ASP D 609 161.312 175.850 111.994 1.00 97.73 O ---------------------------- ~30 A2 difference out of 60-90 A2 ---------- ATOM 19872 CG ASP D 609 160.261 175.386 111.504 1.00 66.78 C ATOM 19873 OD1 ASP D 609 159.374 176.102 110.995 1.00 71.45 O ATOM 19874 OD2 ASP D 609 161.312 175.850 111.994 1.00 64.91 O 2) I found also the default "rotamers.restraints.enabled=true" could really shift the rotamer's refinement adversely. It easily drops the favored rotamer percentage of a good model (refined or manually built) from above 90%s down to 70%s, and likely could never be refined back even with hard endeavors until it was turned off. This is very different experience from that in phenix.refine -- mostly it helped (at least do not set back largely if maps looked good) in the latter. 3) The "rotamers.fit=outliers_or_poormap", upon my tests, is another interesting default? The "outliers_and_poormap" option mostly worked much better for me, while the former one can shift large regional chains a lot (like push long chain out of map; sounds like much over weighted for their contribution?). This even drove me think if there exists a confusion between Boolean and natural languages for this default? 4) Does the default "weight=None" mean this was turned off/NA currently, and all weights were set automatically.... I tested with different values but did not make changes (but BFs as mentioned in 1) above). The instruction says it could be invoked... Best regards, Lijun