On Thu, 2 Oct 2014 11:38:08 +0100, Phil Evans <p...@mrc-lmb.cam.ac.uk> wrote:
>How does XDS decide on eg P 21 21 2 when say c > b > a? The initial indexing >may decide that the cell fits a primitive orthorhombic system, but I presume >that it will then have some convention, probably a < b < c, since the >identification of screws can only be done after integration, and even then may >be uncertain. If e.g. Pointless later decides that the a axis is a dyad, and >the b & c axes are 2(1) screws, then it can assign space group P 2 21 21 >without permuting the indices. If XDS is assigning space group P 21 21 2 then >it must have permuted the axes from the initial indexing. It seems to me more >straightforward to stick to the initial indexing rather than having to reindex >after you have decided the true space group: this was Ian Tickle's point and >is also supposedly the official IUCr-approved convention. XDS was written for users who read the table of H00 0K0 00L intensities and sigmas in CORRECT.LP (i.e. after integration), and know that the only pure two-fold rotation axis should be called c in space group 18, and the only two-fold screw axis should be called c in space group 17. This XDS user then has to choose the correct one out of the three possible ways to order the axes. So, XDS does not decide, the user decides. Nowadays many XDS users use POINTLESS, which is perfectly adequate, except for space groups 17 and 18 where the problem arises, unless the non-default SETTING SYMMETRY-BASED is chosen. I don't see any "sticking to initial indexing" as worthwhile to worry about, since in the first integration, P1 is often used anyway, and it is quite normal (and easy) to re-index after the intensities become available, during scaling. Re-indexing from P1 to the true spacegroup often changes the cell parameters and their order, and this is sufficiently easy and well-documented in the output. > >There are of course ambiguous cases e.g. a ~= b, but that is the same as the >indexing ambiguities in e.g. P3, and that needs a reference dataset to resolve. > >There is no problem in solving a structure in e.g. P 2 21 21, and indeed I >would always run MR in all 8 possible primitive orthorhombic space groups, >very easy to do in Phaser this is true; running in all 8 possible primitive orthorhombic space groups is a fallback that should save the user, and I don't know why it didn't work out in that specific case. Still, personally I find it much cleaner to use the space group number and space group symbol from ITC together with the proper ordering of cell parameters. I rather like to think once about the proper ordering, than to artificially impose a<b<c , and additionally having to specify which is the pure rotation (in 18) or the screw (in 17). And having to specify one out of 1017 / 2017 / 1018/ 2018/ 3018 is super-ugly because a) there is no way I could remember which is which, b) they are not in the ITC, c) XDS and maybe other programs do not understand them. best, Kay > >Phil > >On Tue, 30 Sep 2014 13:29:02 +0100, Phil Evans <p...@mrc-lmb.cam.ac.uk> wrote: > >> Be careful: the International Tables space group number may be ambiguous. >> For example sg number 18 may refer to P 21 21 2 or its permuted settings P >> 21 2 21 or P 2 21 21, if you follow the "proper" IUCr convention that >> primitive orthorhombic space groups have a<b<c > >I would like to point out that there is an alternative interpretation of the >International Tables (Vol A, 4th ed. 1995). In that interpretation (which e.g. >XDS follows) space group 18 has the 'standard' space group symbol, "P21 21 2" >(bold letters in Table 3.2). This is of course not ambiguous at all; the pure >2-fold then corresponds to the "c" axis and there is always a permuation of >axes to achieve this. As a result, the axes are not necessarily ordered such >that a<b<c . The latter ordering is just a "convention" which was "chosen for >convenience" and the "convention refer(s) to the cell obtained by the >transformations from Table 9.3.1" (citing from table 9.3.2) - in other words, >the convention is fulfilled _after_ the transformation (which of course is >just order-permuting while keeping right-handedness) - nothing new here. > >In my understanding, CCP4 developers have (years ago) understood this >"convention" as a "condition", which lead them to invent "CCP4 space group >symbols" 1017 and 2017 as well as 1018, 2018, 3018. This also seems to be the >reason for the default being "SETTING CELL-BASED" in POINTLESS. > >Users of XDS should be aware that by default, POINTLESS therefore permutes the >axes such that a<b<c . This however may lead to space groups 1017 / 2017 / >1018/ 2018/ 3018 - indicated in the MTZ file, but not in the POINTLESS log >file (last I checked). > >In consequence, XDS will use the space group 17 or 18 (which is what POINTLESS >reports), but the user must provide the correct ordering (which does not >necessarily mean a<b<c) of cell parameters in XDS.INP. The easiest way, for >XDS users, would be to run POINTLESS with the "SETTING SYMMETRY-BASED" option >(I wish the latter were the default because the default SETTING CELL-BASED has >no advantages that I can see). Or they use the "good old manual way" of >inspecting, by eye, the systematic absences along H00 0K0 00L - this cannot >fail. > >To me, "symmetry trumps cell metric" so "SETTING SYMMETRY-BASED" should be the >default. > >I'm harping on this because I have recently seen how a Molecular Replacement >solution was not obtained in space group 18 because of the misleading (I'd >say) ordering a<b<c . > >I'm probably also harping on this because it took me so many years to discover >this failure mode, and I would like to prevent others from falling into this >trap. > >HTH, > >Kay > > > >> >> The space group names are unambiguous (though also watch out for R3 & R32 >> which are normally indexed as centred hexagonal, but could be indexed in a >> primitive cell) >> >> Phil >> >> >> On 30 Sep 2014, at 13:07, Simon Kolstoe <simon.kols...@port.ac.uk> wrote: >> >>> Dear ccp4bb, >>> >>> Could someone either provide, or point me to, a list of space-groups >>> relevant to protein crystallography just by space group number? I can find >>> lots of tables that list them by crystal system, lattice etc. but no simple >>> list of numbers. >>> >>> Thanks, >>> >>> Simon
