[ccp4bb] Low resolution, twinning, pseudo-translation
Dear all, I am trying to solve the structure of a protein at ~3 A. Though the structure is already solved in a different space group. I thought it would be a straight forward MR problem. But there are a couple of problems: The possible point group is P3 (2 molecules / asu) and pointless and phenix.xtriage predicted it to be P321 ( 1 molecule /asu). But it shows 37% twinning with P3 whereas twinning is 0.055 % in P321. Several attempts to get a sensible MR solution, in either of the space groups, failed. R and Rfree were always in the range of 35-40 %. In both the cases there is a pseudotranslation of ~17%. Even different runs of the same inputs gave different solutions on multiple trials. After many trials, molrep gave one solution in P3 with 2 molecules/asu which when refined in Phaser with twinning corrections gave R and Rfree of 0.27 and 0.33 respectively. The map looks good and there is nothing really to do in terms of model-building. While my attempts to get better dataset are ON, I was wondering how to ensure that the solution/symmetry which I got are correct (or find out that they are wrong.) and also what next with this data sets in either cases to reach to a correct model? Any suggestion would be highly appreciated. Thanks Ashu
Re: [ccp4bb] Low resolution, twinning, pseudo-translation
First - you probably have the best solution you can hope for - Rs of 27% 33% are not great, but unlikely if you had a wrong answer. A query though: t is surprising to have a pseudo translation if you expect one molecule in the asymmetry unit .. an you elborate a bit on that? Is there a good reason to try to get more information from this 3A twinned data, if the same structure is already solved. If there is a good reason then it is worth persisting, but it will be difficult Eleanor On 5 September 2013 13:09, Ashu Kumar kumar1980a...@gmail.com wrote: Dear all, I am trying to solve the structure of a protein at ~3 A. Though the structure is already solved in a different space group. I thought it would be a straight forward MR problem. But there are a couple of problems: The possible point group is P3 (2 molecules / asu) and pointless and phenix.xtriage predicted it to be P321 ( 1 molecule /asu). But it shows 37% twinning with P3 whereas twinning is 0.055 % in P321. Several attempts to get a sensible MR solution, in either of the space groups, failed. R and Rfree were always in the range of 35-40 %. In both the cases there is a pseudotranslation of ~17%. Even different runs of the same inputs gave different solutions on multiple trials. After many trials, molrep gave one solution in P3 with 2 molecules/asu which when refined in Phaser with twinning corrections gave R and Rfree of 0.27 and 0.33 respectively. The map looks good and there is nothing really to do in terms of model-building. While my attempts to get better dataset are ON, I was wondering how to ensure that the solution/symmetry which I got are correct (or find out that they are wrong.) and also what next with this data sets in either cases to reach to a correct model? Any suggestion would be highly appreciated. Thanks Ashu
Re: [ccp4bb] Low resolution, twinning, pseudo-translation
At the very least look at unit cell packing. Usually the correct space group among possible choices is obvious from crystal packing. Phaser does a good job of placing solutions even with twinned data. A good alternative is EPMR, but we haven't needed to use it in a long time. We had a similar low res case that was either P622 or some flavor of twinned P6 or P3. Only P312 showed reasonable crystal packing, and that solution refined to the mid 20s at 3.0 A Roger Rowlett On Sep 5, 2013 8:19 AM, Ashu Kumar kumar1980a...@gmail.com wrote: Dear all, I am trying to solve the structure of a protein at ~3 A. Though the structure is already solved in a different space group. I thought it would be a straight forward MR problem. But there are a couple of problems: The possible point group is P3 (2 molecules / asu) and pointless and phenix.xtriage predicted it to be P321 ( 1 molecule /asu). But it shows 37% twinning with P3 whereas twinning is 0.055 % in P321. Several attempts to get a sensible MR solution, in either of the space groups, failed. R and Rfree were always in the range of 35-40 %. In both the cases there is a pseudotranslation of ~17%. Even different runs of the same inputs gave different solutions on multiple trials. After many trials, molrep gave one solution in P3 with 2 molecules/asu which when refined in Phaser with twinning corrections gave R and Rfree of 0.27 and 0.33 respectively. The map looks good and there is nothing really to do in terms of model-building. While my attempts to get better dataset are ON, I was wondering how to ensure that the solution/symmetry which I got are correct (or find out that they are wrong.) and also what next with this data sets in either cases to reach to a correct model? Any suggestion would be highly appreciated. Thanks Ashu
Re: [ccp4bb] Low resolution, twinning, pseudo-translation
More Details: This particular crystallization condition (particularly pH) is very differnt from the one in which the first structure was solved. With the analysis of the earlier structure we expected it to acquire a different packing and that is what it appears to take place. So with this low resolution data we are essentially looking at the packing and the assembly. Though I am trying to get better crystals, by the time I get them (if I ) I thought of working on this data to extract and learn as much as possible. I was initially bit confused with the possibility of pseudo-translation when only 1 molecule in the asu was possible. But what I saw in several trials of MR, many times the solutions from different runs were simply related by just tranlations. The molecule is a helical bundle with an elongated shape. To start with, data has good Rmerge ~ 12% only in P3 and P321. In other cases it was around 50%. With Phaser I could get solutions only in P3 and P321 with 2 and 1 copes respectively. In both the cases the packing looked fine, there were no clashes (except terminal 1-2 residues), there were sensible inter-molecular contacts and clear solvent channels were seen. But the map was not good. So I started trying looping one MR output to another run, examining packing and map and so on...With this hit-and-trial method I reached to this particular packing with R and Rfree of 27% and 33%. Since the approach was just hit-and-trial, without any proper scientific steps involved in it, I am bit worried about the authenticity of this solution. Also the 2 molecules which I got are related by a rotation of ~179 degrees and are laterally shifted wrt to each other. Could this information be of any use? I have not tried EPMR yet, I'll give that a try. Thanks a lot On Thu, Sep 5, 2013 at 8:41 AM, Roger Rowlett rrowl...@colgate.edu wrote: At the very least look at unit cell packing. Usually the correct space group among possible choices is obvious from crystal packing. Phaser does a good job of placing solutions even with twinned data. A good alternative is EPMR, but we haven't needed to use it in a long time. We had a similar low res case that was either P622 or some flavor of twinned P6 or P3. Only P312 showed reasonable crystal packing, and that solution refined to the mid 20s at 3.0 A Roger Rowlett On Sep 5, 2013 8:19 AM, Ashu Kumar kumar1980a...@gmail.com wrote: Dear all, I am trying to solve the structure of a protein at ~3 A. Though the structure is already solved in a different space group. I thought it would be a straight forward MR problem. But there are a couple of problems: The possible point group is P3 (2 molecules / asu) and pointless and phenix.xtriage predicted it to be P321 ( 1 molecule /asu). But it shows 37% twinning with P3 whereas twinning is 0.055 % in P321. Several attempts to get a sensible MR solution, in either of the space groups, failed. R and Rfree were always in the range of 35-40 %. In both the cases there is a pseudotranslation of ~17%. Even different runs of the same inputs gave different solutions on multiple trials. After many trials, molrep gave one solution in P3 with 2 molecules/asu which when refined in Phaser with twinning corrections gave R and Rfree of 0.27 and 0.33 respectively. The map looks good and there is nothing really to do in terms of model-building. While my attempts to get better dataset are ON, I was wondering how to ensure that the solution/symmetry which I got are correct (or find out that they are wrong.) and also what next with this data sets in either cases to reach to a correct model? Any suggestion would be highly appreciated. Thanks Ashu
