[buster-discuss] Maintaining ligand planarity during refinement
ashley.pike at sgc.ox.ac.uk
Mon Feb 5 10:27:05 CET 2018
Andrew - Many thanks for looking into this. Marco's solution - tightening global planar restraints with GeometryWeight_plane=20.0 - does indeed do the job but at the expense of overall model geometry (Molprobity is not as pretty). I agree that this solution is far from ideal. Is there no way to apply tighter planar geometry restraints to just ligand?
The density of the base does look a bit non-planar but we can't justify this given limited resolution - the planarity should be strictly maintained. Modelling is not an issue - there are no particularly close contacts - I am about to deposit the model so it really is finalised. I just wanted to sort out this issue w/ ligand as it just looks bad/incorrect. I have tried re-regularising the ligand prior to refinement and making slight adjustments to environment but the structure converges back to the same strained conformation after re-refinement.
I can send you model/data/cif etc. off list if that is helpful.
From: Andrew Sharff [mailto:asharff at globalphasing.com]
Sent: 02 February 2018 14:40
To: Ashley Pike; buster-discuss at globalphasing.com<mailto:buster-discuss at globalphasing.com>
Subject: Re: [buster-discuss] Maintaining ligand planarity during refinement
Thank you for bringing this issue to our attention. We appreciate getting feedback / problem reports from users so we can ensure we continue to improve our software.
Your findings are a little unexpected but we have occasionally had reports about similar behaviour before.
Firstly, it is not at all surprising that you get behaviour closer to your expectation from a restraint dictionary from Grade using the -bigplanes options. By default, Grade generates a series of overlapping 4-atom plane restraints (as you will see if you look at the standard Grade UD1 dictionary). There is also a good reason for doing it this way, since some (large) ring systems actually show real deviation from exact planarity (e.g. FAD) and by using a collection of small, overlapping planes we can handle this. Usually, this approach is quite sufficient to maintain planarity in rings. However, not always. The -bigplanes option sees overlapped 4-atom planes conflated (where appropriate) into a single, large plane. In this case, if you look at the bigplanes dictionary you will see a single plane restraint for the uracyl ring, including C1.
We have tried running BUSTER, with UD1 dictionaries generated a) with Grade, b) with Grade and -bigplanes and c) by CCP4, on a number of PDB structures that contain UD1 and thus far we have been unable to reproduce your problem. Indeed, in one of the models we looked at, 4gx5 (a 3.1A structure) the uracyl ring is non-planar in the deposited model but ends up completely planar after BUSTER (regardless of which dictionary I use).
The suggestion made to you by Marco Bellinzoni may well work but it is an approach that we would urge caution against using, at least at this stage.
We wonder if there is a more particular issue with the local environment around the UD1 that results in the C1 being forced out of plane of the uracyl ring? If the cause is some modelling problem, Marco's suggestion may lead to the planarity being maintained but may simply mask the root cause. His suggestion will also apply globally over the whole structure, not locally on the UD1.
What is the environment around the UD1? Are there any close contacts / residues that need to be resolved? What about alternate conformations?
Also, and despite the resolution, is your model (or just the UD1) hydrogenated - this may also have also have an impact on the refinement. Note that BUSTER does not use a "riding hydrogen" model as refmac does, it includes them in calculation of the geometric terms, so these could well be having an impact.
Would it also be possible for you to send us the model / mtz file so we can look at this directly.
Andrew and Clemens
On 31/01/18 14:41, Ashley Pike wrote:
Does anyone have any tips to force buster to maintain planarity when refining ligands. I am refining a UDP-GlcNAc (UD1) and the N1 of the uracyl base is refining out of plane. Using the bigplanes option when generating the cif helps a lot but the N atom is approx. 0.1A out of the plane (compared to ideal or refmac refined). N1 is in same plane as aromatic ring but not with C1 of ribose. How to I increase weighting towards geometry for ligand - perhaps adding something to gelly file to modify overall weight?
Using the latest snapshot version of buster
Resolution is 3.1A so perhaps I can't expect anything better.
FWIW the cif file used was generated by grade from the PDB ligand entry. The cif file itself is fine as using to refine in coot or refmac results in maintenance of planarity.
Lower buster-refined model shows N1 of uracyl ring out-of-plane wrt to C1 atom of sugar on left-hand side (top models are coot-idealised and refmac-refined)
[cid:image001.png at 01D39C44.AB729A00]
Dr. Ashley Pike
Membrane Protein Crystallography | Structural Genomics Consortium | University of Oxford
buster-discuss mailing list
buster-discuss at globalphasing.com<mailto:buster-discuss at globalphasing.com>
Dr. Andrew J. Sharff D.Phil
Research Scientist / Software Developer
Global Phasing Ltd
Cambridge CB3 0AX
Tel: +(0)1223 353033
Fax: +(0)1223 366889
-------------- next part --------------
An HTML attachment was scrubbed...
-------------- next part --------------
A non-text attachment was scrubbed...
Size: 9478 bytes
More information about the buster-discuss