AMBER Archive (2006)

Subject: Re: AMBER: Mass weights

From: Sergey Samsonov (
Date: Tue Nov 07 2006 - 02:06:14 CST

Thank you very much for the explanation, Mr Cheatham!

The thing I was a bit confused about (and now it's clear for me) is that
I thought that the value one obtains from atomicfluct is also squared
Now I checked that they are not: B-factor = 8/3 * pi *<U>^2. In manual
it is written 'if the keyword "bfactor" is specified, the data is
output as B-factors
rather than atomic positional fluctuations (which simply means
multiplying the results by (8/3)pi**2)'. So it was a bit misleading. Now
I checked both outputs
and it's clear that one parameter is squared derived from the other one.


Thomas Cheatham wrote:
>> I've been using ptraj command atomicfluct. As it is implemented in the
>> algorithm, options bymask and byres of this command dump the results in
>> mass-weighted manner. I would appreciate it very much if someone could give me
>> a reference to the reason/s for the use of mass weighted values for the groups
>> of the atoms. Any information/reference will be enormously appreciated.
> Given that no one else is responded, I provide my rational; but wonder
> up front why we should calculate these values without mass weighting?
> I implemented this originally, based on other codes and common practice,
> to allow comparison of observed atomic fluctuations to those measured
> experimentally. [Note that if you look back in the archives, you can see
> a critical error that has since been resolved in that calculated B-factors
> are mean squared displacements, whereas atomic positional fluctuations are
> not squared (previously both were not squared, hence ptraj calculated
> B-factors were too large).]
> Comparing to experiment, the fluctuations are mass-weighted. As the
> method is extended to residues or groups of atoms, it is not obvious what
> the convention should be (as the values are no longer directly comparable
> to experiment). However, if we assume that "by residue" fluctuations from
> experiment are agregates of atomic (mass weighted) fluctionations,
> weighting by the mass seems appropriate so this is how I implemented it.
> As I do not have a better answer, to delve further, you will need to dig
> into the primary literature or hope for further responses here. Note that
> the ptraj code (transformAtomicFluct() in actions.c) could be modified
> to not do the mass weighting or to have the option...
> -----------------------------------------------------------------------
> The AMBER Mail Reflector
> To post, send mail to
> To unsubscribe, send "unsubscribe amber" to

The AMBER Mail Reflector
To post, send mail to
To unsubscribe, send "unsubscribe amber" to