Hi Dr. Merritt, Thank you very much for your reply.
>> We are trying to solve a protein structure by molecular replacement. >> The protein crystal has unit cell dimensions of a=50 b=51 c=252, >> alpha=90 beta=90.4 gamma=90. > >> We processed the data both as P21 (two molecules in A. U.) > > OK > >> and P21212 (one molecule in A. U.). > > What about the other possible orthorhombic spacegroups? > You need to try at least > P212121 > P22121 > P21221 > You should at the least try that 3rd one, since it is the equivalent > to your P21 setting. I am planning to do that. Did you notice that the model molecule is fatter than my unit cell but we still be able to obtain good density for one molecule in the P21 setting? I can not reprocess the data to check the systematic absence to determine the space group (I do not have the right software to reprocess the raw image at this time). But P21 and P21212 gave Rmerge around 10% up to 2.5 A resolution. I checked with the lab members, about three postdoc fellows worked on this project before, I was told the space group was sure P21212. >> We have built a model based on sequence homology. >> It is very interesting that we >> still found one molecule using P21 data, reasonably GOOD density >> although the model is about 60 A wide and 82 A long (just refine this >> one molecule using P21 data, the Overall R factor=0.4731, Overall >> figure of merit=0.4939, the protein molecule seems have no symmetry by >> itself). > > Wait. > So your P21 solution only contains one molecule of the two that are expected? > If you can't fit a second molecule in P21, then I don't see how moving to > a higher-symmetry space group can work. I am sorry my message may not that clear. We failed P21212 then tried P21. >> We can not find any solution using P21212. The protein >> seems have three domains based on our model. We tried to using only >> one of the domains. The results were similar. Any suggestions as to >> what else is possibly wrong except that we have not found the correct >> solution? > > I suggest not to use a homology model for molecular replacement. > Better to use the homologous structure[s] directly, without modifying > their coordinates. See recent thread. Thanks for the good suggestions. Chunmin
