Saturday, 21 January 2012

refmac Bsol


From: Bernhard Rupp (Hofkristallrat a.D.)
Date: 29 December 2011 21:08

Dear Refmac developers group,

I am trying to understand a small detail in the refmac log listing. In the bulk

solvent section just below  the restraint table (I use monitor MANY in general):

-----------------------------------------------------------------------------

Overall               : scale =    0.891, B  =  -0.266

Partial structure    1: scale =    0.375, B  =  24.388

Overall anisotropic scale factors

   B11 = -0.39 B22 =  0.09 B33 =  0.33 B12 = -0.00 B13 =  0.48 B23 =  0.00

-----------------------------------------------------------------------------

Is the listing indeed containing the flat bulk solvent model ('simple model' in refmac gui, SOLVENT YES) k and B in the first 2 lines?

If so, I am curious

a)      what exactly is 'Overall' versus 'partial structure' What is the relative magnitude of the scale and B telling me in each case?

b)       would it not be useful to have it reported in the PDB header?

      All I find there is the anisotropic B scaling matrix and a mean (overall) B value (of unspecified origin how exactly is that computed?)

      REMARK   3  B VALUES.

      REMARK   3   FROM WILSON PLOT           (A**2) : NULL

      REMARK   3   MEAN B VALUE      (OVERALL, A**2) :  18.711

Explanation/intuition would be much appreciated.

Best regards, BR

-----------------------------------------------------------------

Bernhard Rupp

------------------------------------------------------------------

Old and treacherous will beat young and skilled every time

------------------------------------------------------------------

PS: The doc leaves me a little confused because SCAL type SIMP would imply

KB = K0*exp(-B0*s^2)  (Simple Wilson scaling)

i.e. K1 = 0

while BULK (Babinet) which I did not specify

KB = K0*exp(-B0*s^2) * (1- K1*exp(-B1*s^2))

gives me 2 Ks and Bs.

So I conclude that 'overall' and 'partial' lines do  not list Babinet ks and Bs (the B is also too small for that)


----------
From: Garib N Murshudov


Partial structure mask bulk solvent parameters. If you use simple scaling then mask solvent is still on. You can turn it off by "Calculate the contribution of from the solvent region"

To turn Babient's bulk solvent you need to use Babinet's scaling instead of simple.

I hope it helps

regards
Garib
Garib N Murshudov 




----------
From: Bernhard Rupp (Hofkristallrat a.D.)


Dear Garib,

 

thank you for the quick response despite (or because of) the holidays. I'll try to summarize because I am not sure I understand yet, and it might be useful for all :

 

Ø  Partial structure mask bulk solvent parameters. Mask bulk solvent B value is in addition to protein B value.

 

Not quite clear what you are saying here – how does that answer what the 'Overall' means versus the 'Partial'? (some guessing below)

 

Ø  Full scaling is like this:

 

Fscaled = scale_protein exp(-B s^2/4) exp(-s^T U s) (F_protein + scale_mask exp(-B_mask s^2/4) Fmask) (1-scale_babinet exp(-B_babinet s^2/4))

 

Ok then I see that I can use either _mask or _babinet terms because only one set of terms is in effect if the other one is zero. If I turn both off, I get basic k, B scaling, plus anisotropic B correction. Makes perfect sense.

 

Based on that I will try to determine which  line in the printout is actually the bulk solvent contribution. For the 'Partial structure' part:

If I recall correctly, for the flat bulk solvent model the scale_mask and  B_mask are generally in the order of 0.4 and 40AA, so then this ought to be scale_mask and  B_mask of the solvent contribution? Yes? No? Now for

Now it gets interesting: what is this in terms of the above equation? I don't seem to be able to factor out a single overall scale and B correction from your eqn, but it sure looks like a correction term from bulk to be applied to the overall k and B…. Can you clarify please?

Ø  Current version of refmac does not calculate Wilson B value

I think we can get that readily from say truncate. But that won't be in the PDB REMARK 3 then…

 

Ø  Note: Overall Bvalue may correspond to residual overall B value if you are using TLS refinement

 

Understood.

 

Ø  If you use simple scaling then mask solvent is still on.

 

Yes – that is consistent with the doc.

Ø  You can turn it off by "Calculate the contribution of from the solvent region"

 

Yes, that can be done by unclicking the box "Calculate the contribution from the solvent region". Handy if one needs to check certain things…

 

Ø  To turn Babient's bulk solvent you need to use Babinet's scaling instead of simple.

 

Yep, understood.

 

Thx, BR



----------
From: Garib N Murshudov

Dear Berhnard

It seems there are only one or two questions remain to clarify.

Overall in the pdb header is calculated from atominc B values.
Partial is mask bulk solvent. 
Again: Overall B value in the pdb header is from atominc B values. Bvalue here is remaining part. It should be 0 but because of differences between lsq scaling and ML it is not always 0. For refinenemnt we use ML and for scaling/Rfactor calculations ls scaling. 
Yes. It would not be in the pdb header after refmac. It needs to be sorted out. I.e. refmac should calculate/read and write tot the pdb header
Regards and Happy New Year
Garib


No comments:

Post a Comment