[phenixbb] Modeling sugars with Coot and Phenix

Emilia C. Arturo (Emily) ecgarturo at gmail.com
Sat Jan 2 17:51:50 PST 2021

Hello All,

I have several questions regarding sugar modeling. These are questions for
both Coot (0.9.4-pre/CCPEM) and Phenix (1.18.2-3874) usage (on a Mac) for
modeling as correctly as possible several N-linked glycosylation (i.e. not
ligand) sugars within a protein, and about the approach as it pertains to
higher resolution X-ray maps (better than 2A) or to lower resolution EM
maps (~3-4A). I am new to the world of glycoproteins and cryoEM, not new to
model building or X-ray crystallography, and would welcome any advice or

For relatively low resolution maps, where sugar moieties cannot be placed
unambiguously, in theory (say Agirre  (2017) and Emsley and Crispin (2018))
pyranose placement within the map should be quite highly restrained by
ideal geometry for the specific sugar expected from prior knowledge. In
practice, however, it’s not entirely clear how to maintain these restraints
on pyranose geometry and conformation during manual building in Coot, or
how to apply any special restraints in Phenix. The Glyco module of Coot
does indeed place fine looking Asn-NAGs after a few iterations of manual
placing and refining to account for the poor resolution of the acetyl
group, just as discussed by Emsley and Crispin (2018). However, the
validation report that is obtained from the OneDep system at the rcsb web
server for the resulting structure shows many outliers in bonds and angles
within the NAG or between the NAG and protein Asn residue. In Coot I have
experimented with various restraints (in the R/RC menu), including
decreasing the Refinement weight/weight matrix to zero, or some small
non-zero number, thinking this places less weight on the map signal and
more on the ideal values within Coot’s libraries. I have tried a variety of
refinement restraint combinations in Coot to address individual bonds and
angles outliers within the sugar, but the rcsb validation report still
finds departures from ideal geometry. Meanwhile, in Phenix I don’t do
anything specific for the refinement of glycosylation sugars (should I?)
other than apply the appropriate restraints for any low resolution EM map;
are there specific restraints files I ought to be using within Phenix real
space refine jobs for N-linked sugars?

Then I noticed that loads of existing PDB entries containing N-linked
glycans contain similar outliers in pyranose geometry. So this prompted me
to wonder whether we’re not doing things right, that in practice we’re not
correctly following what’s advised in theory, or whether this is just the
state of model building currently. Could someone more experienced than am I
on sugar building please comment here on how exactly to get the sugars
better built? e.g. what buttons to press and why, in Phenix and/or in Coot.
Or is this a tolerable situation where sugar geometries differ from ideal
even at low resolution? Is it perhaps a matter of different targets being
used for PDB validation by the rcsb versus by Coot?

It turns out that regarding higher resolution maps I am also unsure of
exactly how to apply restraints within Coot. For example, using PDB ID 5VAA
and its map with a reported resolution of 1.55A, I centered on Asn297 and
its associated glycosylation sugars. I experimented with refinement
strategies (i.e. the settings of R/RC), thinking that the best approach
would be little restraint on ideal geometry and higher weight on the map
signal. But no matter what combination of restraints I used, I was unable
to keep the model locally from distorting; i.e. the deposited model versus
map looks great, but within Coot I find a hundred ways to destroy it. I
must be doing something wrong.

Any advice or references you’d recommend would be greatly appreciated.


"Study as if you were going to live forever; live as if you were going to
die tomorrow." - Maria Mitchell

"I'm not afraid of storms for I'm learning to sail my ship."  - Louisa May
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://phenix-online.org/pipermail/phenixbb/attachments/20210102/750698f1/attachment.htm>

More information about the phenixbb mailing list