To judge the significance of the value of a bond length, you cannot rely on global measures of precision. You need the standard uncertainty of this particular distance in this particular crystal. Refmac will not perform this calculation for you. Shelxl will. It is certainly more cumbersome to use, but it will tell you the estimate for the length of this bond along with the standard uncertainty which you can use to judge the statistical significance of the deviation.
Your choice of restraints will be difficult. Since you believe that the C=O bond might be partially broken, that electron has to go somewhere and some other bond might get stronger. I would recommend applying restraints to the protein and all the parts of the NAD except the catalytic part. Removing these restraints will increase your SU's but decrease any systematic errors due to incorrect library values. With 1.1A data the calculation might be stable. Dale Tronrud Ivan Shabalin wrote: > Dear all, > > I have an atomic resolution (1.1Å) structure of enzyme with the bound > cofactor NAD. During the analysis of the refined structure I found that > important double C=O bond of the cofactor in the active site was slightly > lengthened from standard 1.22Å to 1.26Å. Then I increased the > sigma's of the > distance restraint in the REFMAC dictionary 50-fold and as a result the > distance increased to 1.29 Å. I suggested that this deviation from > standard > double C=O bond means decrease in the bond order to something between 1 and > 2. This change in bond order upon cofactor binding (and especially in the > transition state of the reaction) was suggested previously by studies of the > kinetic isotope effects and partly confirmed by quantum mechanics > calculations. > > This change in 0.07Å is quite small, but is a bit higher than the DPI > multiplied by a factor of two (2*0.028=0.056). Is it crystallographically > sound to state the bond lengthening? Is DPI the correct estimation of error > of positional parameters in my case (note that B-factors in the active site > are about 12-15, when average B-factor for all atoms is 20)? Maybe, ML based > su of positional parameters is better estimation. > > Statistics of the dataset (values in parentheses are for the > highest-resolution shell 1.11-1.10 Å): > R-merge = 5.1 (52)% > Redundancy = 4.0 (2.5), > Completeness = 96.5 (73.8)% > <I>/<σ(I)> =44 (1.9), > mosaicity = 0.7 (I think, this value is a little bit high). > > Statistics of the refinement with REFMAC: > R-factor =13.4% > R-free =15.9%, > RMSD_BOND = 0.015Å > DPI = 0.028 > ML based su of positional parameters = 0.0217. > Estimated with SFCHECK maximal error 0.062 A > Estimated with SFCHECK minimal error 0.014 A > > All help greatly appreciated!! > > Yours, > Ivan
