> As you see, distortion index table indicates I centered tetragonal, I > centered orthorhombic, F centered orthorhombic, C centered monoclinic > and triclinic as possible Bravais lattices. > > Data processed in I centered tetragonal gives low Rmerge in all the > possible space groups namely I4, I41, I422 and even I4122. Other > space groups in lower symmetry lattices also gave low R merge values > (around 6% in most of the cases).
If it processes nicely in I422, it will also in subgroups of I422 (sub as I4, I222, various choices of C2 and of course P1). > Since we have not been able to obtain a solution in any of the space > group from I centered tetragonal to triclinic (I4, I4122, I222, C2 and > even P1) using Se-SAD, we decided to check the data for any intrinsic > problem such as twinning. Do you have any anomalous signal? <snap> > > Now my question is: > - Are data showing low Rmerge value in I4122 due to nearly perfect > twin in space group I4? > That is possible, but seems unlikely. > - Why cumulative intensity distribution shows a normal pattern for > the data where as Yeates server and SFcheck indicates nearly a perfect > twin? Why Yeates server and SFCheck shows different twin fraction and > twin operator? Twin fraction estimates rely on correlations between twin related intensities. If you process yuor data in a spacegroup that is too low, you will see that 1. You have twin laws available for your crystal symmetry 2. You have strong correlations between twin related reflections In both cases, the cause is the that you have wrong symmetry. Twinning analyses should take into account the presence of wrong symmetry. If that is not done, question like this come up. Furthermore, the yeates server unfortunately doesn't deal with pseudo merohedral twinning. The differences in twin operators is due to the crystal symmetry: the operators listed are related by the symmetry of I4. The difference in twin fraction is due to difference s in data cuts and resolution limits I suppose. I highly recommend using phenix.xtriage (for obvious reasons) to give you a relatively clear picture of what is going on. Get it from the latest phenix release. usage is straightforward. > - Is it possible to detwin this data and use it for structure solution? I guess you first want to make sure that your experimental data has any anomalous signal at all before your blame twinning to be the reason that you cannot solve your structure. HTH P
