Temp Fact Variance Analysis

From: Uma Ratu
Date: 1 March 2012 14:26

Hello,

 

I run my model in Coot to do "Temp Fact Variance Analysis".

There are red bars from the B-factor Variance graphy.

I click each red bar to exam the residues in Coot. Many of these residues do not have the electronic density on their side chains, especially Lys residues.

 

Here is my questions:

 

1. The lack of electronic density is the cause of these red-bars?

2. How do I fix them? delete the side chains?

 

The diffraction data is 2A, and data completness is 98.8%.

 

Thank you for comments

 

Ros

 

----------
From: Kelly Daughtry


Ros,

I haven't used the  "Temp Fact Variance Analysis" in Coot, but can guess that the red bars indicate increased b-factor compared to the average of your protein model?

If so, then:

1. The lack of electronic density is the cause of these red-bars?

Likely yes. If there is no density to support the position of the side chain, then you will see an increase in B-factor.

2. How do I fix them? delete the side chains?

This question has been under debate in our community recently.

If I recall correctly (and my opinion) is you should leave the side chains as they are and let the b-factor be an indication of the uncertainty of their position. 

I'm willing to bet these side chains are surface residues, or in a mobile part of our protein.

A good thing to compare is the average b-factor of your entire model compared to these residues, as well as the b-factor of these specific side chains compared to  main chain atoms.

Hope this helps clarify.

Sincerely, 

Kelly Daughtry

*******************************************************
Kelly Daughtry, Ph.D.

----------
From: Tim Gruene 

Dear Ros,

yes, the lack of electron density is most likely the cause the these red
bars (unless you set the map level rather high, but at the default
chosen by coot that should be fine).

yes - if you don't see electron density you don't have experimental
evidence of the atom positions, so it is best to delete them.

Tim

- --
- --
Dr Tim Gruene

----------
From: Ed Pozharski 

> 2. How do I fix them? delete the side chains?

Here we go again. Take a look at these threads

http://www.dl.ac.uk/list-archive-public/ccp4bb/msg19738.html
http://phenix-online.org/pipermail/phenixbb/2011-March/016875.html


--
Oh, suddenly throwing a giraffe into a volcano to make water is crazy?
                                               Julian, King of Lemurs

----------
From: <Herman.Schreuder


Here we go again!

We neither have evidence that some mysterious side-chainnase has chopped
off all flexible surface side chains and in fact, at least I am pretty
sure these side chains are still there and moving around in the solvent
as indicated by their very/extremely high B-factors, which, again for
me, is the better description of the "real" situation.

My 2 cents!
Herman

----------
From: Vellieux Frederic


Hi,

You could try to lower the threshold used to contour the maps for these side-chains (middle mouse "button" if you have a 3 button mouse with a wheel in the middle). Quite often such side chains have lowish electron density that does not appear at (say) 1.0 or 1.5 sigma (in the 2mFo-DFc maps), but that appears at (say) 0.7 sigma. If you delete such side-chain atoms you'll probably end up with negative difference density appearing in the difference Fourier after the next round of refinement. So these residues must be checked individually, much work I know...

"To delete or not to delete, that is the question".

Fred.

----------
From: Vellieux Frederic


Typo (my fingers type faster than my brain...)

--


Fred,

After deleting the difference density will be positive, IMHO.

   Vaheh

----------
From: Uma Ratu

Dear All:

 

Thank you very much for your comments.

 

I did not notice that this issue has been dicussed lately.

 

Thank you for your inputs

 

regards

 

Ros

----------
From: Paul Emsley

I'd think twice about "fixing" these red bars.

The signal to noise in the temperature factor variance graphs are unimpressively low.  Currently red bars indicate "there might be something wrong here" not "there is something wrong here".

Paul.

----------
From: Eleanor Dodson


Hard to know for sure - I often set occ to 0.0, then redo refinement and maps, only to find they have reappeared in the density at a low level.. In the end all our models are defective - LYS and ARg and GLU etc often have alternate conformations which we cant model at low resolution, and the temperature factor gives an indication that that is so..

But remember - any atom with B = 60 is going to be more spread out than one with B = 20, and you need to drop the contour level when you assess the high B factor structure..

Eleanor