Hello Lisa--

> 
> Any official conclusion on this RDC data / Cartesian minimization issue?  Is
> it best for those of us using RDC data to skip that final Cartesian space
> minimization? Or maybe to perform it but without the RDC potential?
> 

The thread was rather confused, and without proper resolution. The
initial message was regarding using RDCs to try to refine a
(presumably low resolution) crystal structure:

> I had near perfect correlation between observed and calculated RDCs
> values, which was fishy. After addition of 5Hz systematic noise to all
> RDCs, i still had perfect correlation and nice convergence with no
> violations(!). Initially i thought that there was something wrong with
> my experimental setup (since i only used RDCs+dihedrals restrains),
> but I had exactly the same result when I changed also the gb1 tutorial
> RDCs values.
> So to sum it up, I used the tutorial script (refine.py) and input
> files for RDC refinement, regardless of how much I alter the RDCs
> values (i tried up to +6Hz noise addition) I get always nice
> convergence, perfect correlation obs/calc with no violations.
> 

In this case the crystal structure was represented by dihedral
restraints, and presumably these restraints were not violated when
RDCs were added and subsequently overfit. The real question here is:
does the addition of RDCs improve the structure when used in this
fashion. Alternative approaches of refining the crystal structure
would be 1) to use NCS-style restraints (using posDiffPot) or (best!)
2) use the X-ray data along with RDCs in a refinement. With either of
these approaches RDC overfitting should be much less of an issue- but
still something to watch for, and perhaps tuning the energy scale
factors might be necessary in such protocols. This message got lost in
the discussion.

The thread continued with a suggestion that the default RDC energy
scale factors in the eginput/gb1_rdc/refine.py script might be too
large, but evidence really wasn't presented for that: the RDCs might
be overfit, but that should clearly be indicated by violations or poor
fit in other terms.

Finally, there was a suggestion that final Cartesian minimization was
somehow messing things up. Again, one must be careful- particularly of
angle violations when many RDCs are used - but this step is actually
useful in getting rid of bad contacts, so I would still recommend it.

Other opinions welcome--
Charles
_______________________________________________
Xplor-nih mailing list
[email protected]
https://dcb.cit.nih.gov/mailman/listinfo/xplor-nih

Reply via email to