Yes, you can do better.
Look at your structure and its electron density in Coot. Walk through
the structure by clicking the space bar, and use the validation tools.
You 'll notice several pepflips, missing solvent with 2 or 3 clearly
visible hydrogen bonds, flipped histidines/ asparagines etc. Don't rely
so much on automated methods and don't drown your structure with water.
Automated waters tend to hide possible alternative (or main-)
conformations from view. Start by refining your protein structure and
add only those waters that fit in a network with multiple hydrogen bonds
and make chemical sense.
Polishing with a bucket full is only for the last stage, if ever.
Don't focus on R-factors, your refinement program knows about them.
Focus on the chemistry.
cheers,
Hans
Lucas schreef op 2016-09-26 20:05:
> I've followed with great attention the thread about 5ggn, and it
> reminded me of a structure I solved eight years ago. It was actually
> crystallized long before that, but the people working on the lab back
> then failed to solve the structure due to the lack of a search model
> for MR (there was no available homolog at the time) and also the fact
> there were no softwares to detect twinning back then. Better armed
> back in 2008, I managed to go on, found a solution using MOLREP with
> 2HYK (released a couple of years before) as a search model.
>
> XTRIAGE detected pseudomerohedral twinning, and so I remember doing
> everything I could to get that done right, including looking for
> similar cases in the literature, and found out two articles (Declercq,
> J. P. and Evrard, C. (2001) Acta Cryst., D57, 1829. and Rudolph, M.
> G., Wingren, C., Crowley, M. P., Chien, Y.-H. and Wilson, I.A. (2004)
> Acta Cryst., D60, 656.) on which the authors describes this case on
> twinning in a monoclinic system when fortuitously the cell has
> c*cos(beta)=-a/2.
>
> So I proceeded with refinement, and the result is 3ILN, which we
> finally published at J. Phys Chem B three years later (it also
> included molecular dynamics simulation and discussion on
> thermostability/thermophilicity). The current version at PDBREDO (
> http://www.cmbi.ru.nl/pdb_redo/il/3iln/index.html ) made some
> improvements to the structure we deposited back then, but I wonder if
> there were any other serious error that went unnoticed, because it
> does have a pretty crappy clashscore. Could a similar (I hope not as
> serious) error have been done in that structure?
|