[ccp4bb] AW: [ccp4bb] Spacegroups, screw axes and ordering
Dear Kai, I agree that Ian's response was strongly worded, but I don't think it is personal. I agree with Ian that assignment of screw axes based on axial reflections is very unreliable. That is why I always advice people to run molecular replacement in all possible space groups, independent of what the data processing programs may claim. If a data set gets reindexed because molecular replacement found out that the initial assignment of screw axes was incorrect, this is also a great source of confusion, especially if novices follow the advice of the bulletin board to always use the original mtz file as input for refinement! Best, Herman -Ursprüngliche Nachricht- Von: CCP4 bulletin board [mailto:CCP4BB@JISCMAIL.AC.UK] Im Auftrag von Kay Diederichs Gesendet: Freitag, 29. Januar 2016 13:51 An: CCP4BB@JISCMAIL.AC.UK Betreff: Re: [ccp4bb] Spacegroups, screw axes and ordering Ian, we have not met in the past and did not have any scientific conflict, AFAIK. Thus, I am really surprised that you attack my opinion so verbosely, with strong words such as "seriously misrepresenting" and "asserting". Maybe a cultural or language difference of some sort that I have not so far encountered, nor do I understand it. Anyway, I won't try to answer in the same or similar way. sincerely, Kay -- Kay Diederichshttp://strucbio.biologie.uni-konstanz.de email: kay.diederi...@uni-konstanz.deTel +49 7531 88 4049 Fax 3183 Fachbereich Biologie, Universität Konstanz, Box M647, D-78457 Konstanz
[ccp4bb] AW: AW: [ccp4bb] Spacegroups, screw axes and ordering
Dear Kai, you are right, but the original question was posed in the context of xia2. At that stage I think only diffraction data are available to decide about screw axes. Best, Herman -Ursprüngliche Nachricht- Von: Kay Diederichs [mailto:kay.diederi...@uni-konstanz.de] Gesendet: Freitag, 29. Januar 2016 16:16 An: Schreuder, Herman R&D/DE; CCP4BB@JISCMAIL.AC.UK Betreff: Re: AW: [ccp4bb] Spacegroups, screw axes and ordering Dear Herman, I did not write anything about the question of assigning a spacegroup based on screw axes absences. _Of course_ this is unreliable, but this was not the question (to me, at least). And of course I agree that one should try molecular replacement with all possible space groups, unless the assignment is clear. The question, as I understand it, was whether _if you know where the screw axis is and where not_ if you then want a) to order the axes such that a < b < c , and call the space group e.g. P 21 2 21 (because in that axis ordering, it is the b axis that does not have a screw), or, b) if you want to have the "unique" axis last, giving P 2 2 21 or P 21 21 2 , respectively, and accept the fact that this may result in a > b or b > c . And my preference is choice b) for reasons I tried to explain. best, Kay On 2016-01-29 15:09, herman.schreu...@sanofi.com wrote: > Dear Kai, > > I agree that Ian's response was strongly worded, but I don't think it is > personal. I agree with Ian that assignment of screw axes based on axial > reflections is very unreliable. That is why I always advice people to run > molecular replacement in all possible space groups, independent of what the > data processing programs may claim. If a data set gets reindexed because > molecular replacement found out that the initial assignment of screw axes was > incorrect, this is also a great source of confusion, especially if novices > follow the advice of the bulletin board to always use the original mtz file > as input for refinement! > > Best, > Herman > > -Ursprüngliche Nachricht- > Von: CCP4 bulletin board [mailto:CCP4BB@JISCMAIL.AC.UK] Im Auftrag von Kay > Diederichs > Gesendet: Freitag, 29. Januar 2016 13:51 > An: CCP4BB@JISCMAIL.AC.UK > Betreff: Re: [ccp4bb] Spacegroups, screw axes and ordering > > Ian, > > we have not met in the past and did not have any scientific conflict, AFAIK. > > Thus, I am really surprised that you attack my opinion so verbosely, > with strong words such as "seriously misrepresenting" and "asserting". > Maybe a cultural or language difference of some sort that I have not so > far encountered, nor do I understand it. > > Anyway, I won't try to answer in the same or similar way. > > sincerely, > Kay > -- Kay Diederichshttp://strucbio.biologie.uni-konstanz.de email: kay.diederi...@uni-konstanz.deTel +49 7531 88 4049 Fax 3183 Fachbereich Biologie, Universität Konstanz, Box M647, D-78457 Konstanz This e-mail is digitally signed. If your e-mail client does not have the necessary capabilities, just ignore the attached signature "smime.p7s".
Re: [ccp4bb] Spacegroups, screw axes and ordering
Not sure I fully understand. Are we really talking about non-crystallographer scientists, often willing to understand how a biologically meaningful molecule / entity looks like? Are these non-crystallographers really interested in crystal packing issues? Is it so much difficult to press on a couple of buttons in the program with which you do open your coordinates to generate the symmetry related molecules? I am feeling we are a bit off here... Leo > On 29 Jan 2016, at 20:52, Quyen Hoang wrote: > > Sure, but you would need to expand space group X in order to see the > intermolecular interactions that would have been seen in P1. Also, it is > often discussed here about how non-crystallographers might use our structural > models deposited in the PDB, I doubt that many of them would know how to > expand. > > I am not advocating, just discussing. > > Cheers, > Quyen > > Sent from my BlackBerry 10 smartphone. > From: Keller, Jacob > Sent: Friday, January 29, 2016 2:20 PM > To: Quyen Hoang; CCP4BB@JISCMAIL.AC.UK > Subject: RE: [ccp4bb] Spacegroups, screw axes and ordering > > >I mean, would it not be more informative to have a P1 unit cell filled with > >molecules compared to a single molecule representing only a fraction of the > >unit cell? > > No, it would not be more informative: a model in space group X can easily be > expanded to P1. > > JPK > > > Cheers, > Quyen > > On Jan 29, 2016, at 12:18 PM, Keller, Jacob wrote: > > >Sure, but wouldn’t the same could be achieved by NCS averaging? > > Yes, with complete “NCS” constraints it would be the same. But why do P1 if > so—you’d have all the same issues when deciding which parts of the “NCS” to > constrain (it would be tantamount to SG determination.) Using unmerged data, > however, would have some advantages (one could model the variations between > reflections in a more direct way.) > > I guess the final goal would be to reproduce as accurately as possible the > diffraction images. Thus, crystallographic refinement becomes data-faking (I > guess all science is this same data-faking, in a way.) > > JPK > > > > Ed, regarding the fractional problem with molecular replacement and data to > parameter ratio problem in refinement, I am sure that you know how to get > around these problems ;) > > Quyen > > On Jan 29, 2016, at 10:41 AM, Bernie wrote: > > Precision is always better when averaging is applied. Mirror planes and > rotations will be perfect rather than exact within experimental error. There > are also singularities in the refinement process that can force the structure > to be symmetrically imperfect. > > Sent from my iPhone > > On Jan 29, 2016, at 8:10 AM, Quyen Hoang wrote: > > Given enough data and modern computing powers, why don’t we just use P1? > > Quyen > > On Jan 29, 2016, at 8:45 AM, George Sheldrick > wrote: > > The collection and scaling requires the Laue group but not the space group. > For small molecule structure determination, many more space groups have to be > considered and the choice may be ambiguous (like I222 and I212121) or > difficult, so my current small molecule structure solution program SHELXT > only uses the input space group to deduce the Laue group. After solving the > structure with the data expanded to P1 it uses the phasesto determine the > space group and origin shift. In practice this is much more reliable than > using systematic absences. This was not my idea (see papers by Giacovazzo and > Palatinus), I just wrote a little program to implement it. How the user has > chosen a, b and c is irrelevant, the program outputs the solution in the > conventional setting for the space group in question (as the correct > enantiomorph based on the Friedel differences where appropriate). It also > finds the most compact arrangement of atoms and centers it is the unit-cell. > > Whether this was worth the effort is debatable. SHELXT has been freely > available for the last couple of years, but the open access paper that > explains how it works (Acta A71 (2015) 3-8) is rarely cited. > > George > > > On 01/29/2016 01:06 PM, Ian Tickle wrote: > > Hi Kay > > You are seriously misrepresenting how this works in practice. Isomorphism > always takes precedence over convention: convention is not an absolute > requirement! Convention is the _default_ in the absence of all other > criteria (that's why we have conventions!). Only the _first_ crystal in an > isomorphous series would be indexed by convention, the others would be > indexed using that as a reference (i.e. based on the intensity correlation, > _not_ the unit cell or the assumed space group which may not be reliable, > using REFINDEX, which is what we have always used, or POINTLESS) - very > simple! At Astex we have be doing this for our large fragment screens for 15 > years with no problem. > > In any case we find that assignment of screw axes by axial reflexions is very > u
Re: [ccp4bb] Spacegroups, screw axes and ordering
Sorry if I don't buy it. Just personal feeling. I don't see how someone capable of navigating through the coordinates, see meaningful interactions, highlight those and understand them, would not be able to press these buttons. While some programs indeed do not have these options (which makes me wonder about the usefulness of these programs by the way), most of those would. Let me join also Jacob's comment on the non usefulness of P1 in here, and send these non-experts either to PISA or to other experts more knowledgeable, say the people who solved these structures for instance. Your point on the mtz file is interesting. We are now moving things toward some non-crystallographers in need to install quite a few programs, if not at least libraries, and put in these mtz files enough information to reproduce all the non-standard refinement procedures experts do apply for their structure refinement. Not to mention 5A structures for which auto build might be trickier to run than building in 1A maps. Leo > On 29 Jan 2016, at 22:02, Quyen Hoang wrote: > > No, I wasn’t talking about crystal packing, I was thinking about potential > homo oligomeric interactions that might be important for function. > If we are talking about the simplicity of pushing couple buttons and saving > disk space and bandwidth, then I guess the same argument could be made that a > simple MTZ file containing phases should suffice? > The non-crystallographer would simply press a few buttons to generate a model > with auto build? > But I have a feeling those couple of buttons might not be so obvious to > people in other fields. > > Quyen > >> On Jan 29, 2016, at 3:16 PM, CHAVAS Leonard >> wrote: >> >> Not sure I fully understand. Are we really talking about >> non-crystallographer scientists, often willing to understand how a >> biologically meaningful molecule / entity looks like? Are these >> non-crystallographers really interested in crystal packing issues? Is it so >> much difficult to press on a couple of buttons in the program with which you >> do open your coordinates to generate the symmetry related molecules? I am >> feeling we are a bit off here... >> >> Leo >> >>> On 29 Jan 2016, at 20:52, Quyen Hoang wrote: >>> >>> Sure, but you would need to expand space group X in order to see the >>> intermolecular interactions that would have been seen in P1. Also, it is >>> often discussed here about how non-crystallographers might use our >>> structural models deposited in the PDB, I doubt that many of them would >>> know how to expand. >>> >>> I am not advocating, just discussing. >>> >>> Cheers, >>> Quyen >>> >>> Sent from my BlackBerry 10 smartphone. >>> From: Keller, Jacob >>> Sent: Friday, January 29, 2016 2:20 PM >>> To: Quyen Hoang; CCP4BB@JISCMAIL.AC.UK >>> Subject: RE: [ccp4bb] Spacegroups, screw axes and ordering >>> I mean, would it not be more informative to have a P1 unit cell filled with molecules compared to a single molecule representing only a fraction of the unit cell? >>> >>> No, it would not be more informative: a model in space group X can easily >>> be expanded to P1. >>> >>> JPK >>> >>> >>> Cheers, >>> Quyen >>> >>> On Jan 29, 2016, at 12:18 PM, Keller, Jacob >>> wrote: >>> Sure, but wouldn’t the same could be achieved by NCS averaging? >>> >>> Yes, with complete “NCS” constraints it would be the same. But why do P1 if >>> so—you’d have all the same issues when deciding which parts of the “NCS” to >>> constrain (it would be tantamount to SG determination.) Using unmerged >>> data, however, would have some advantages (one could model the variations >>> between reflections in a more direct way.) >>> >>> I guess the final goal would be to reproduce as accurately as possible the >>> diffraction images. Thus, crystallographic refinement becomes data-faking >>> (I guess all science is this same data-faking, in a way.) >>> >>> JPK >>> >>> >>> >>> Ed, regarding the fractional problem with molecular replacement and data to >>> parameter ratio problem in refinement, I am sure that you know how to get >>> around these problems ;) >>> >>> Quyen >>> >>> On Jan 29, 2016, at 10:41 AM, Bernie wrote: >>> >>> Precision is always better when averaging is applied. Mirror planes and >>> rotations will be perfect rather than exact within experimental error. >>> There are also singularities in the refinement process that can force the >>> structure to be symmetrically imperfect. >>> >>> Sent from my iPhone >>> >>> On Jan 29, 2016, at 8:10 AM, Quyen Hoang wrote: >>> >>> Given enough data and modern computing powers, why don’t we just use P1? >>> >>> Quyen >>> >>> On Jan 29, 2016, at 8:45 AM, George Sheldrick >>> wrote: >>> >>> The collection and scaling requires the Laue group but not the space group. >>> For small molecule structure determination, many more space groups have to >>> be considered and the choice may be ambiguous (like
Re: [ccp4bb] Spacegroups, screw axes and ordering
>No, I wasn’t talking about crystal packing, I was thinking about potential >homo oligomeric interactions that might be important for function. If we are talking about the simplicity of pushing couple buttons and saving disk space and bandwidth, then I guess the same argument could be made that a simple MTZ file containing phases should suffice? The non-crystallographer would simply press a few buttons to generate a model with auto build? But I have a feeling those couple of buttons might not be so obvious to people in other fields. Maybe just the number "42" would suffice? JPK Quyen > On Jan 29, 2016, at 3:16 PM, CHAVAS Leonard > wrote: > > Not sure I fully understand. Are we really talking about non-crystallographer > scientists, often willing to understand how a biologically meaningful > molecule / entity looks like? Are these non-crystallographers really > interested in crystal packing issues? Is it so much difficult to press on a > couple of buttons in the program with which you do open your coordinates to > generate the symmetry related molecules? I am feeling we are a bit off here... > > Leo > >> On 29 Jan 2016, at 20:52, Quyen Hoang wrote: >> >> Sure, but you would need to expand space group X in order to see the >> intermolecular interactions that would have been seen in P1. Also, it is >> often discussed here about how non-crystallographers might use our >> structural models deposited in the PDB, I doubt that many of them would >> know how to expand. >> >> I am not advocating, just discussing. >> >> Cheers, >> Quyen >> >> Sent from my BlackBerry 10 smartphone. >> From: Keller, Jacob >> Sent: Friday, January 29, 2016 2:20 PM >> To: Quyen Hoang; CCP4BB@JISCMAIL.AC.UK >> Subject: RE: [ccp4bb] Spacegroups, screw axes and ordering >> >>> I mean, would it not be more informative to have a P1 unit cell filled with >>> molecules compared to a single molecule representing only a fraction of the >>> unit cell? >> >> No, it would not be more informative: a model in space group X can easily be >> expanded to P1. >> >> JPK >> >> >> Cheers, >> Quyen >> >> On Jan 29, 2016, at 12:18 PM, Keller, Jacob wrote: >> >>> Sure, but wouldn’t the same could be achieved by NCS averaging? >> >> Yes, with complete “NCS” constraints it would be the same. But why do P1 if >> so—you’d have all the same issues when deciding which parts of the “NCS” to >> constrain (it would be tantamount to SG determination.) Using unmerged data, >> however, would have some advantages (one could model the variations between >> reflections in a more direct way.) >> >> I guess the final goal would be to reproduce as accurately as possible the >> diffraction images. Thus, crystallographic refinement becomes data-faking (I >> guess all science is this same data-faking, in a way.) >> >> JPK >> >> >> >> Ed, regarding the fractional problem with molecular replacement and data to >> parameter ratio problem in refinement, I am sure that you know how to get >> around these problems ;) >> >> Quyen >> >> On Jan 29, 2016, at 10:41 AM, Bernie wrote: >> >> Precision is always better when averaging is applied. Mirror planes and >> rotations will be perfect rather than exact within experimental error. There >> are also singularities in the refinement process that can force the >> structure to be symmetrically imperfect. >> >> Sent from my iPhone >> >> On Jan 29, 2016, at 8:10 AM, Quyen Hoang wrote: >> >> Given enough data and modern computing powers, why don’t we just use P1? >> >> Quyen >> >> On Jan 29, 2016, at 8:45 AM, George Sheldrick >> wrote: >> >> The collection and scaling requires the Laue group but not the space group. >> For small molecule structure determination, many more space groups have to >> be considered and the choice may be ambiguous (like I222 and I212121) or >> difficult, so my current small molecule structure solution program SHELXT >> only uses the input space group to deduce the Laue group. After solving the >> structure with the data expanded to P1 it uses the phasesto determine the >> space group and origin shift. In practice this is much more reliable than >> using systematic absences. This was not my idea (see papers by Giacovazzo >> and Palatinus), I just wrote a little program to implement it. How the user >> has chosen a, b and c is irrelevant, the program outputs the solution in the >> conventional setting for the space group in question (as the correct >> enantiomorph based on the Friedel differences where appropriate). It also >> finds the most compact arrangement of atoms and centers it is the unit-cell. >> >> Whether this was worth the effort is debatable. SHELXT has been freely >> available for the last couple of years, but the open access paper that >> explains how it works (Acta A71 (2015) 3-8) is rarely cited. >> >> George >> >> >> On 01/29/2016 01:06 PM, Ian Tickle wrote: >> >> Hi Kay >> >> You are seriously misrepresenting
Re: [ccp4bb] Spacegroups, screw axes and ordering
I guess again it depends on people and the procedure that scientists do adopt during their analysis. Yes, you could have more information in these specific cases in P1. Yet, those are specific cases and what would be the frequency of such happening? Biologically speaking, do we have so many homo-oligomers? I am not molecular biologist so cannot really comment on those, however I doubt it would justify putting everything in P1. You can take it the other way around then: for monomeric proteins, you would need to go down individual molecules and ignore these interactions. Pushing things a bit: you could even have referees asking for proving these interactions they see in the unit cell between molecules are not biologically meaningful etc. I am certainly pushing things a bit, but we do have a good working system and I don't feel strong enough science cases to force changing it. Saying that, I am a drop in the ocean of crystallographers, with what people might classify as zero knowledge in biology, just liking to argue things ^^ Leo > On 29 Jan 2016, at 22:46, Quyen Hoang wrote: > > Let get back to science for a moment, assuming that there is enough data (as > stated in my original post about this) would you not agree that the models > build into a P1 space group would represent the content of the unit cell more > accurately than a higher space group (albeit could be insignificant)? > After your refinement has converged, don’t you always expand to look at the > packing and potential oligomeric states of your molecule? Let's say that you > have a true tetramer, what would be lost if you report your structural model > in P1 instead of P212121? > I mean, is there a theoretical or scientific reason against reporting P1 for > a crystal system consisting of higher symmetry? > > Cheers, > Quyen > >> On Jan 29, 2016, at 4:27 PM, CHAVAS Leonard >> wrote: >> >> Sorry if I don't buy it. Just personal feeling. I don't see how someone >> capable of navigating through the coordinates, see meaningful interactions, >> highlight those and understand them, would not be able to press these >> buttons. While some programs indeed do not have these options (which makes >> me wonder about the usefulness of these programs by the way), most of those >> would. Let me join also Jacob's comment on the non usefulness of P1 in here, >> and send these non-experts either to PISA or to other experts more >> knowledgeable, say the people who solved these structures for instance. >> >> Your point on the mtz file is interesting. We are now moving things toward >> some non-crystallographers in need to install quite a few programs, if not >> at least libraries, and put in these mtz files enough information to >> reproduce all the non-standard refinement procedures experts do apply for >> their structure refinement. Not to mention 5A structures for which auto >> build might be trickier to run than building in 1A maps. >> >> Leo >> >>> On 29 Jan 2016, at 22:02, Quyen Hoang wrote: >>> >>> No, I wasn’t talking about crystal packing, I was thinking about potential >>> homo oligomeric interactions that might be important for function. >>> If we are talking about the simplicity of pushing couple buttons and saving >>> disk space and bandwidth, then I guess the same argument could be made that >>> a simple MTZ file containing phases should suffice? >>> The non-crystallographer would simply press a few buttons to generate a >>> model with auto build? >>> But I have a feeling those couple of buttons might not be so obvious to >>> people in other fields. >>> >>> Quyen >>> On Jan 29, 2016, at 3:16 PM, CHAVAS Leonard wrote: Not sure I fully understand. Are we really talking about non-crystallographer scientists, often willing to understand how a biologically meaningful molecule / entity looks like? Are these non-crystallographers really interested in crystal packing issues? Is it so much difficult to press on a couple of buttons in the program with which you do open your coordinates to generate the symmetry related molecules? I am feeling we are a bit off here... Leo > On 29 Jan 2016, at 20:52, Quyen Hoang wrote: > > Sure, but you would need to expand space group X in order to see the > intermolecular interactions that would have been seen in P1. Also, it is > often discussed here about how non-crystallographers might use our > structural models deposited in the PDB, I doubt that many of them would > know how to expand. > > I am not advocating, just discussing. > > Cheers, > Quyen > > Sent from my BlackBerry 10 smartphone. > From: Keller, Jacob > Sent: Friday, January 29, 2016 2:20 PM > To: Quyen Hoang; CCP4BB@JISCMAIL.AC.UK > Subject: RE: [ccp4bb] Spacegroups, screw axes and ordering > >> I mean, would it not be more informative to have a P
