From: Uma Ratu
Date: 7 March 2012 16:20
----------
From: Roger Rowlett
----------
From: Uma Ratu
----------
From: Joel Tyndall
----------
From: Uma Ratu
----------
From: Bernhard Rupp (Hofkristallrat a.D.)
----------
From: Eleanor Dodson
I guess we would all like to be able to do that!
High resolution structures show that a) there are well defined H2Os with tidy H bonds. b) there are multiple networks where the waters (and many side chains) have partial occupancy c) there is a soup of other "stuff" which was in the crystallisation medium and is very difficult to model..
There are some lessons - never forget what was in the crystallisation ..
remember that the solvent structure will never be complete
good luck
eleanor
--
Professor Eleanor Dodson
----------
From: Parthasarathy Sampathkumar
Dear Uma,
----------
From: Shekhar Mande
Well....just to add, it has been our contention that many of the metal ions have been modelled as waters in several structures- due perhaps to the lack of sufficiently high resolution data. We published some of the potential metal binding sites in many structures a few years ago:
Proteins. 2008 Mar;70(4):1206-18.
Shekhar--
----------
From: Uma Ratu
Date: 7 March 2012 16:20
Dear All:
I try to add water to my model.
Here is how I did:
Coot: Find Wates
Map: FWT PHWT; 1.8 rmsd; Distances to protein atoms: 2.4 min/3.2 max
Coot found 270 water molecules.
I then examed these waters. Most of them had ball shape. Some had two or more balls together. Some had irregular shape (not glabol shape).
I run Water Check. The program did not find any mis-matched water.
Here is my question: how could I tell the waters are real? Or something else?
Thank you for advice
Ros
----------
From: Roger Rowlett
Uma,
Remember that your structure, ultimately, is a model. A model is your best judgment of the true representation of the protein structure in your crystal. Your model should make chemical sense. Coot is pretty good at placing waters, but it cannot substitute entirely for the experimentalist. Coot will miss some waters, and mis-assign others into weak, unmodeled or alternate side- or main-chain density, or into density that might be attributable to cations and anions or other crystallization materials. Your waters should be subjected to inspection and verification. It is really helpful to turn on environment distances in Coot when you do this. Even in a large protein model, it is possible to inspect all waters for reasonableness pretty quickly. If you have no significant positive or negative difference density, and the b-factors are not way out of line, and hydrogen bonding partners are reasonable, then modeling a water is probably a good call.
Waters should have hydrogen bonding partners with side chains or main-chain polar atoms, within reasonable distances, or be within hydrogen bonding distance of other waters that are (chains of waters). If a "water" has strong electron density and more than 4 polar contacts, you might consider anion or cation occupancy. Most anions and cations will have higher electron density, and appropriately different types of polar contacts. (e.g. you might find sulfates near a cluster of basic residues). Low occupancy anions can often look a lot like water. PEGs can create ugly "snakes" of variable density that may be challenging to model. Modeling non-protein structural bits is endlessly entertaining for the protein crystallographer. ;)
Cheers,
____
Remember that your structure, ultimately, is a model. A model is your best judgment of the true representation of the protein structure in your crystal. Your model should make chemical sense. Coot is pretty good at placing waters, but it cannot substitute entirely for the experimentalist. Coot will miss some waters, and mis-assign others into weak, unmodeled or alternate side- or main-chain density, or into density that might be attributable to cations and anions or other crystallization materials. Your waters should be subjected to inspection and verification. It is really helpful to turn on environment distances in Coot when you do this. Even in a large protein model, it is possible to inspect all waters for reasonableness pretty quickly. If you have no significant positive or negative difference density, and the b-factors are not way out of line, and hydrogen bonding partners are reasonable, then modeling a water is probably a good call.
Waters should have hydrogen bonding partners with side chains or main-chain polar atoms, within reasonable distances, or be within hydrogen bonding distance of other waters that are (chains of waters). If a "water" has strong electron density and more than 4 polar contacts, you might consider anion or cation occupancy. Most anions and cations will have higher electron density, and appropriately different types of polar contacts. (e.g. you might find sulfates near a cluster of basic residues). Low occupancy anions can often look a lot like water. PEGs can create ugly "snakes" of variable density that may be challenging to model. Modeling non-protein structural bits is endlessly entertaining for the protein crystallographer. ;)
Cheers,
____
----------
From: Uma Ratu
Dear Roger:
Thank you very much for your comments. I use them as guideline and remove many 'false waters".
Still, I am not clear of some of these 'waters' are real or not. I have the pic attached.
In Pic-W11-1, the 'water' is connected to the adjust residues with 4 contacts, which are 'N' or 'O' atoms. I would consider this 'water' is false. My question is: if these 4 contacts include "C" from residues, will it be a polar contact or not?
In Pic-W12-1, the 'water' is connected to the adjust residues with 3 contacts. The 4th is to another 'water'.
Will this 'water' is true or not? Similar case is seen in Pic-W190-1
In Pic-W109-1, some 'waters' are connected to adjust residues, some not. Are these 'water' true or not?
Further more,
> and the b-factors are not way out of line,
I am not clear on how to define "out of line".
How to find b-factor of individual residue in Coot? I search the web, but find no answer.
Thank you for advice
Uma
----------
From: Joel Tyndall
Hi Uma,
Water has the capability of making 4 h-bonds, 2 from the two non-bonding pairs of electrons (h-bond acceptors - expect an N-H from an amide for example) as well as the two hydrogens (h-bond donors). I would refine all those waters and assume they are waters. If the distance to the other atoms is between 2.5-3.2 then you can assume the water to be correct. In many cases waters will h-bond (only) to other water molecules.
The B-factor is displayed in Coot along the bottom (left) when you middle click on an atom. You can also see the B-factor when you read the pdb file as text
Hope this helps.
_________________________________
Joel Tyndall, PhD
----------
From: Uma Ratu
Hi, Joel:
Thank you for your comments.
Some of these waters have 4 contacts all with "O" from adjacent residues. As "O" can be doner as well as acceptor, I would consider them as 'real' water. Some of these 'water' have more than 4 contacts, I would consider them as 'false'.
I lower the H-bond seeting to 2 - 3.2. This helps to reduce the noise.
>The B-factor is displayed in Coot along the bottom (left) when you middle click on an atom. You can also see the B-factor when you read the pdb file as text
I found the B-factor using both ways.
Thank you very much!
Uma
----------
From: Bernhard Rupp (Hofkristallrat a.D.)
Ø Some of these 'water' have more than 4 contacts, I would consider them as 'false'.
How about bifurcated hydrogen bonds?
BR
----------
From: Eleanor Dodson
I guess we would all like to be able to do that!
High resolution structures show that a) there are well defined H2Os with tidy H bonds. b) there are multiple networks where the waters (and many side chains) have partial occupancy c) there is a soup of other "stuff" which was in the crystallisation medium and is very difficult to model..
There are some lessons - never forget what was in the crystallisation ..
remember that the solvent structure will never be complete
good luck
eleanor
--
Professor Eleanor Dodson
----------
From: Parthasarathy Sampathkumar
Dear Uma,
The water pictured in W12-1.jpg: could this be a potential metal ion? If you flip the side chain on Asn at 3.08Angstrom, then this has 3 or 4 coordination with oxygen atoms. So, provided your crystallization condition or buffer contains metal ion(s), you could attempt to see if it fits better with a refinement cycle.
May be a similar situation with the water described in W11-1.jpg as well? Difficult to say from these figures.
COOT within the "validate" wizard has an option to search for "hihgly-coordinated waters" like the one you have pictured.
Hope this helps,
Partha
----------
From: Shekhar Mande
Well....just to add, it has been our contention that many of the metal ions have been modelled as waters in several structures- due perhaps to the lack of sufficiently high resolution data. We published some of the potential metal binding sites in many structures a few years ago:
Proteins. 2008 Mar;70(4):1206-18.
Shekhar--
----------
From: Uma Ratu
Dear All:
I appreciate for your comments and inputs.
Thank you
Uma
No comments:
Post a Comment