Dear All, I think Harry raises a valid point in his example of the positional uncertainty of the Ser OG. In this example, the disordered atom can either be anywhere on a circular path or on two or more discrete positions on the same circle. This however is also true for serines modelled in weak density in a single conformation in say a 3Å structure, compared to a similar structure at high resolution. We know that there is a possibility of multiple side chain conformations, but we model the single one that we can see more or less. The uncertainty in this case is never areason for leaving out the OG. Indeed, one crystallographers alternate conformation is the other ones noise... What is more wrong (or less right): modelling the most plausible position of the OG, model multiple positions at random, or just leaving it out? I second Herman's suggestion of loosening up the B-value restraints for atoms on rotatable bonds, this way the problem will pretty much go away by itself. Flip
_____ From: [EMAIL PROTECTED] [mailto:[EMAIL PROTECTED] On Behalf Of Harry M. Greenblatt Sent: Thursday, January 11, 2007 12:42 To: ccp4BB CCP4 Subject: Re: [ccp4bb]: Modelling disordered side-chains BS"D Dear All, This topic has been discussed before, and if I recall correctly, someone pointed out a very important thing about B-factors. Perhaps I missed something, but I have not seen this point raised at this time. Everyone (?) on this list knows the following: B-factors measure displacement about an average position due to thermal vibration. Thus, the bond between atoms should not be regarded as a stick, rather a spring. Because most of us have poor data/parameter ratios, we are forced to model this vibration spherically. Small molecule people have a more accurate ellipsoid model. I think everyone should appreciate the following: Missing side chain density is generally attributed to rotational disorder. Imagine a serine side chain, where there is free rotation about the Chi1 torsional angle. The Og will follow a circular path. Another possibility is that the Og has 2 or more discrete positions, and rotation about Chi1 is not free. Because of lack of resolution, however, we don't see the Og, since it's occupancy at each position is lowered. So how can one consider taking a side chain for which there is no density, presumably because it's moving (or was moving before it was flash cooled...?) or assuming a few discrete conformations, and assign very high thermal vibration parameters in the hope to model this disorder? Seems rather incorrect to me. Leave out the atoms you don't see. As far as non crystallographers not understanding our data: caveat emptor. Harry ------------------------------------------------------------------------- Harry M. Greenblatt Staff Scientist Dept of Structural Biology [EMAIL PROTECTED] Weizmann Institute of Science Phone: 972-8-934-3625 Rehovot, 76100 Facsimile: 972-8-934-4159 Israel
