I wouldn't call it numerical error, as such. Refmac doesn't restrain/constrain the T, L or S values at all. In particular, the L value can refine stably and accurately to a negative value.
Refmac is refining a set of ADPs constrained to fit a certain functional form with parameters T, L and S. That functional form is derived from a rigid body model. If L goes negative, then the assumption of a rigid body breaks down. But you still have a set of ADPs that describe your data. So with an L eigenvalue = -0.268 you can probably say that you are close to a rigid body model with virtually no libration about one axis. The rigid body model is only a rough assumption - you should have the same caveats if you have L = +0.268 With L = -53.975, you may well have numerical instabilities, or you are so far from a rigid body model that the results are meaningless. And your individual ADPs are probably bad as well. Cheers Martyn On Tue, 2007-02-27 at 09:23 -0500, Richard Gillilan wrote: > I failed to mention that this particular group is a small set of > terminal residues (48 atoms I think) that are part of a poorly- > resolved HIS tag, so it turns out not to be a meaningful part of the > model anyway. I was just surprised to see negative values reported. > The other body (different structure) that gives a negative value has > as the smallest mean-squared displacement -0.27. I take it that this > simply means that the mean-squared displacement is near zero and that > I am seeing numerical error. It seems strange to encounter a near- > zero displacement in some direction, but one of the examples in the > TLS documentation shows exactly this behavior as well. Here are the > parameters: > > > AXES OF LIBRATION WRT TO MEAN-SQUARE ANGLE LIBRATION > AXES MAKE TO > ORTHOGONAL AXES (IN ROWS) DISPLACEMENT ORTHOGONAL AXES > (DEG) > ABOUT AXES (DEG^2) X > Y Z > 0.633 0.420 0.650 4.952 50.71 > 65.17 49.45 > -0.363 0.903 -0.230 -0.268 111.29 > 25.44 103.27 > -0.683 -0.091 0.724 10.899 133.12 > 95.21 43.59 > > MEAN LIBRATION (TRACE/3) 5.194 > > On Feb 26, 2007, at 4:40 PM, Winn, MD (Martyn) wrote: > > > I'd also comment that the absolute values are quite large. Well, I > > have no > > knowledge what groups you've chosen, but for domain-sized groups, > > typical > > values are 1-10 deg**2. Values can be larger for smaller TLS > > groups. This is > > just a rule-of-thumb, but would make me suspicious whether the TLS > > refinement > > was really stable in this case. > > > > Regards > > Martyn > > > > -----Original Message----- > > From: CCP4 bulletin board on behalf of Richard Gillilan > > Sent: Sun 2/25/2007 1:10 PM > > To: [email protected] > > Subject: [ccp4bb] TLSANL: negative mean-sq displacements? > > > > After TLS refinement (which seemed to be stable and produced nice > > R_free values), I have analyzed the rigid body results with TLSANL. > > I get negative mean-square displacements along the axes of libration > > WRT to orthogonal axes! Am I misunderstanding something here? The > > units are (deg^2). I see this with two different structures. Here is > > the output from TLSANL: > > > > AXES OF LIBRATION WRT TO MEAN-SQUARE ANGLE LIBRATION AXES > > MAKE TO > > ORTHOGONAL AXES (IN ROWS) DISPLACEMENT ORTHOGONAL AXES (DEG) > > ABOUT AXES (DEG^2) X Y Z > > 0.842 0.197 0.502 -53.975 32.66 78.63 59.85 > > -0.494 0.657 0.570 193.421 119.60 48.97 55.24 > > -0.217 -0.728 0.650 -12.276 102.55 136.73 49.45 > > > > MEAN LIBRATION (TRACE/3) 42.390 > > > > > > Anyone seen this happen before? > > > > Richard Gillilan > > MacCHESS >
