Hi Jorge, I have had some similar cases. The twin fraction can vary during data collection on the synchrotron (different areas of the crystal can have different twin fractions). Sometimes this is not a problem and twin refinement works fine - sometimes it is a problem and it is hard to get a clear electron density map from the structure. R-factors from twinned structures tend to be lower than 'normal' R-factors - your ~19/25% would be OK for normal data but ... If the twin fraction varies as you rotate the crystal and illuminate different volumes you can sometimes select fewer images to help.
Kind regards, Ben On Wed, Jul 20, 2022 at 10:12 PM Kay Diederichs < kay.diederi...@uni-konstanz.de> wrote: > Dear Jorge, > > what you write makes sense to me, and I cannot answer your questions. This > is just to say that the situation you encounter is not completely uncommon, > although most crystallographers would abandon such a crystal form, I guess. > The technical term that describes this 4-fold twinning is "tetartohedral > twinning" (in contrast to merohedral twinning, which involves two twin > domains); maybe this helps to find additional pointers. > > best wishes, > Kay > > On Wed, 20 Jul 2022 11:10:04 -0300, Jorge Iulek <jiu...@gmail.com> wrote: > > >Dear all, > > > > > > We had some data collections at a Synchrotron. Crystals are a kind > >of brush like (lattice strains, to use a term by Dr. Bergfors, though we > >employed good effort for purification), but we took advantage of the > >Synchrotron microfocus. > > > > Some of the datasets (images) clearly shows more than one lattice > >(maybe more than two) that, struggling, we managed to process a get a > >dataset which allowed molecular replacement and then (initial) refinement. > > > > But, in a second Synchrotron travel (and after efforts for > >improving crystals), we got in some cases images with spots "well > >separated' "unique lattice" at some of the target spots (radiation on > >the crystal) in the crystal. > > > > We processed these happily to P212121 (though some strange points > >by pointless and/or xtriage, namely that " the L-test suggests that the > >data may be twinned, so the indicated Laue symmetry may be too high"). > >Systematic absences seem to be OK for lower resolution reflections, but > >at higher resolution there seems to be more of a modulation (if a look > >at a P222 processing). > > > > Anyway, we took, initially, refinement at P212121, nevertheless (I > >should say not surprisingly), it stuck at 30/31 % R-free, model close > >(if not at all) to completion. Data resolution is 2.31 A. > > > > We went to process these images in P1, and in the three possible > >P21 (named P21_45, P21_122 and P21_155 - according to approx. axis > >dimensions) sg's. Then we went to refine (refmac, twin option) the > >current model (and then due "symmetry copies" produced with pdbset and > >added to the model to be refined,) in all possible space groups, and > >*care was taken* to inherit the former r-free set *and* make the then > >corresponding twin related reflections to be in the r-free set (to be > >close to 5% of reflections, but "independent" reflections). It turned > >out that the R/Rfree values dropped around to ~19/25% for P1 and one > >(namely, P21_b151) of the P21; higher values for other P21's. As > >expected, twin domains refine more or less close to 25 (P1) and 50% (any > >P21), respectively. > > > > To mess up a bit more, I made the same study with "another dataset" > >(another illuminated spot on the - same - crystal). In this case, only > >the dataset processed in P1 presented "good" R values. > > > > I think these observations might correlate to what the "crystal " > >physically is... a mix of portions genuinely P212121 but mixed, more or > >less, that in some places with twins in one or more other types, > >depending on where I focus my beam. > > > > Should I look at anything else to establish twin P1 is the best way > >to refine this structure? > > > > > > Related, and a question I mentioned before in this forum: what if a > >genuine 2 axis (say , P222 to P2, or even to P1) (I do not mean this is > >the case here) is ignored such that one would have doubled the number of > >observations but also doubled the number of parameters to be refined > >(suppose to exclude NCS in any case). Would one expect R/Rfree values to > >be similar in both P222 and P2 (or even P1)? How much extra freedom > >would one have besides the twin domain fractions to refine? > > > > > > Yours, > > > > > >Jorge > > > >######################################################################## > > > >To unsubscribe from the CCP4BB list, click the following link: > >https://www.jiscmail.ac.uk/cgi-bin/WA-JISC.exe?SUBED1=CCP4BB&A=1 > > > >This message was issued to members of www.jiscmail.ac.uk/CCP4BB, a > mailing list hosted by www.jiscmail.ac.uk, terms & conditions are > available at https://www.jiscmail.ac.uk/policyandsecurity/ > > ######################################################################## > > To unsubscribe from the CCP4BB list, click the following link: > https://www.jiscmail.ac.uk/cgi-bin/WA-JISC.exe?SUBED1=CCP4BB&A=1 > > This message was issued to members of www.jiscmail.ac.uk/CCP4BB, a > mailing list hosted by www.jiscmail.ac.uk, terms & conditions are > available at https://www.jiscmail.ac.uk/policyandsecurity/ > ######################################################################## To unsubscribe from the CCP4BB list, click the following link: https://www.jiscmail.ac.uk/cgi-bin/WA-JISC.exe?SUBED1=CCP4BB&A=1 This message was issued to members of www.jiscmail.ac.uk/CCP4BB, a mailing list hosted by www.jiscmail.ac.uk, terms & conditions are available at https://www.jiscmail.ac.uk/policyandsecurity/
