Sure, the RDCs should be further scaled inversely proportional to the square of their exp. errors. So you should trust the NH Da/R more than those coming from the NC’, of course. You can scale the NC’ RDCs about say 10 times weaker than the NHs, for example (just by guessing the ballpark relative exp. errors) and see if the ANGL violations decrease. You shouldn’t worry much about very small angle violations, even if they are many, they may be caused by the minimization in the torsion angle coordinate space with different RDC sets that contradict each other to a small degree (this has been discussed in the lit. - see for example Ad Bax’s papers)
Gabriel https://sites.google.com/a/wisc.edu/gabrielc/ > On Jun 4, 2016, at 11:21 AM, Emmanouilidis, Leonidas > <[email protected]> wrote: > > After experimenting a bit: > Dihedral Angle restrains + NC rdcs (let Da and R float)--> good result with > no violations and r-factor 0.14 (Da found 49 and R 0.249) > from rdcPotTools import create_RDCPot, > rdcs = PotList('rdc') > for (medium,expt,file, scale) in \ > [('l','NCO','RDC_nc_tpr_cns_noloops.tab' ,1) > ]: > rdc = create_RDCPot("%s_%s"%(medium,expt),file,media[medium]) > > > Dihedral angle restrains + NH rdcs (let Da and R float again) --> good result > no violations with r-factor for NHs fitting 0.41 (Da found 24 and R 0.4) > from rdcPotTools import create_RDCPot, > rdcs = PotList('rdc') > for (medium,expt,file, scale) in \ > [('l','NH' ,'RDC_hn_tpr_cns_noloops.tab' ,1), > ]: > rdc = create_RDCPot("%s_%s"%(medium,expt),file,media[medium]) > > > But when I use both RDC from NHs and NCs i get many ANGL violations. > > The fact that i get two different Values for Da and R depending on the data > set, should ring any bell on how should i treat them when i combine them? > > thank you > Leonidas > > > > On Jun 3, 2016, at 5:06 PM, Gabriel Cornilescu wrote: > >> Additionally, the relative sign of the RDCs (due to various sign conventions >> for 15N) may get inadvertently flipped, so it’s a good idea to test refining >> with each RDC set separately. Or, you can SVD fit the carbon RDCs them to a >> model structure that fits well the NHs and pay attention to the sign of the >> best fitting alignment tensor axes sign etc. Sometimes empirically flipping >> the sign of the NH RDCs when SVD fitting in conjunction with say CaC would >> result in a much better fit, proving that their (intial) relative sign was >> wrong. >> >> Cheers, >> Gabriel >> >> >> >> >>> On Jun 3, 2016, at 9:32 AM, Charles Schwieters <[email protected] >>> <mailto:[email protected]>> wrote: >>> >>> >>> Hello Leonidas-- >>> >>>> >>>> I am trying to refine a crystal structure with RDCs, including NH, NC, >>>> and CaC (and dihedral angles). While NHs and dihedral angles give nice >>>> result, when I include any of the carbon RDCs the algorithm does not >>>> converge due to many violations. I wonder if there is any method to >>>> weight different thos NC and CaC RDCs to influence less than the NH, >>>> if that makes sense. >>>> >>> >>> 1) Please read the paragraph at the top of >>> https://nmr.cit.nih.gov/xplor-nih/doc/current/python/ref/rdcPotTools.html >>> <https://nmr.cit.nih.gov/xplor-nih/doc/current/python/ref/rdcPotTools.html> >>> about the RDC sign convention. >>> >>> 2) If you prescale the RDCs (don't use scale_toNH), then you usually >>> will need to weight the energy terms differently, but if you do not >>> prescale the RDCs (and thus use scale_toNH in the script), the >>> energies of the RDCs involving different nuclei should scale >>> appropriately (such that signal to noise is approximately the same) >>> without different weight factors. >>> >>> 3) The number of violations is not such a useful concept for RDCs, as >>> the threshold for violation would generally be different for those >>> of large and small magnitude. Thus, we usually use the R-factor (or >>> possibly chi^2) to assess the goodness of RDC fit. >>> >>> I hope this helps-- >>> Charles >>> _______________________________________________ >>> Xplor-nih mailing list >>> [email protected] <mailto:[email protected]> >>> https://dcb.cit.nih.gov/mailman/listinfo/xplor-nih >>> <https://dcb.cit.nih.gov/mailman/listinfo/xplor-nih> >> >
_______________________________________________ Xplor-nih mailing list [email protected] https://dcb.cit.nih.gov/mailman/listinfo/xplor-nih
