negative density in difference map

Careina Edgooms 

Sent: Wednesday, 23 November 2011 6:54 PM

Good morning CCP4 members

I have a question about a 2F0-Fc difference map that I calculated with Refmac. In some instances it gives me negative (red) density around part of a side chain and no positive density in sight. Furthermore the entire residue fits well into the blue density of the complete map, including the part with negative density. 

I am struggling to interpret this. Does the fact that it fits the blue density mean that the side chain is in the correct place or does the red blob on part of the side chain (eg on the sulphur in a Met residue) mean that something funky is happening with this side chain?

Thanks for any assistance

Careina

From: DUFF, Anthony
Date: 23 November 2011 08:39

Delete (set occupancies to zero) the side chain back to CA.  Do a few rounds of refinement and calculate Fo-Fc and examine.

 

It is possible that the side chain is disordered, or ordered in multiple conformations.  Compare the density for alternate confonformers against the density for CA.

 

Alternatively, your side chain B-factor restaints might be too tight.

 

Anthony

 

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

I wish I could answer this!
One possibility is that the side-chain B values are too tightly restrained - Ian Tickle recommends releasing these somewhat..

Here are the default refmac values.
THERMAL FACTORS
         Weight= 1.00
    Main chain bond (1-2 neighbour)          1.5A**2
    Main chain angle (1-3 neighbour)         2.0A**2
    Side chain bond                          3.0A**2
    Side chain angle                         4.5A**2

and you could change them say to:


Under GUI - check geometric parameters and alter the Bfactor values

data line becomes:  temp 1.0     1.5 2.0  4.0 6.0

The trouble is that resyraints for ARG or MET say should be looser than those for SER or VAL.

But I often finish up with some inexplicable red blobs - sometimes floating in space..
Eleanor

----------
From: Ian Tickle


Hi Careina

Since my name came up I felt compelled to comment, though I don't have
a definitive answer.  Over-restraining of B factors is certainly a
possibility and could well explain otherwise inexplicable difference
density.  IMO B factors are generally over-restrained.  They are a bit
like the R factors: tight restraints keep the B factors low and people
(referees in particular!) seem to feel happier when they see low
values, even though that is not necessarily optimal (this mind-set is
of course known as "wishful thinking" - see Wikipedia article).

Personally I never restrain 1-3 B factors, and only use the 1-2
restraints, and even then I relax those from the default values.  The
1-3 B restraints are clearly not independent of the 1-2 ones, and
restraints should ideally be independent of each other, otherwise you
are applying them more than once,  IMO the overall B factor restraint
weight should be optimised in the same way that the X-ray weight
should be, by minimising Rfree or -LLfree.  X-plor, CNS (and I would
guess phenix.refine) have had this option for many years
(optimize_rweight.inp in CNS), and it seems an excellent thing to do,

Disorder is also an obvious possibility as Anthony said, and I would
follow his sensible advice.

Cheers

-- Ian

Cheers

-- Ian