Re: diffusion tensor related -- January 17, 2007 - 03:04

There are two factors which cause what you are seeing.  Firstly there
are inconsistencies between the definitions of theta and phi in
different programs.  In relax theta is the azimuthal angle (the
longitude), but sometimes phi is defined as the azimuthal angle.  Also
relax assumes the azimuthal angle is measured from the x-axis whereas
it sometimes measured from the y-axis.

The second reason is also very important and is an optimisation
problem.  The 'r2r1_diffusion' diffusion program use the R2/R1 ratio
to determine the diffusion tensor.  The problem with this methodology
is that it assumes that are no significant internal motions.  This,
however, is definitely not the case!  Therefore this is a very rough
estimate used as a starting point for model-free optimisation.  This
is the reason why model-free analysis needs to be iterated.  The
diffusion tensor as the start is a poor estimate and after the first
round of optimisation is still distant from the true values.  See the
post by Sebastien Morin
(https://mail.gna.org/public/relax-users/2006-11/msg00015.html) for
information about how long it can take for full convergence of the
diffusion tensor and model-free parameters.  Generally you'll need up
to 5 iterations for the isotropic diffusion of a sphere.  For the
spheroid and ellipsoid (axially symmetric and fully anisotropic
diffusion) you'll need over 10 iterations.  Then, and only then, will
you need to do Monte Carlo simulations.

Edward


On 1/17/07, Hongyan Li <hylichem@xxxxxxxxxxxx> wrote:
> I have tried to use "diffusion_tensor.init(name, (6.77e-9, 0.8, 80, 300),
> param_types=2, spheroid_type='prolate', fixed=1)" in mf-multimodel.py script,
> the tensor was obtained from r2r1_diffusion, I supposed that tensor should be
> fixed, but when I checked results from relax, both theta and phi have been
> changed. Why?
>
> Cheers!
>
> Hongyan
>
> Quoting Chris MacRaild <c.a.macraild@xxxxxxxxxxx>:
>
> > On Fri, 2006-12-29 at 12:12 +0800, Hongyan Li wrote:
> > > Dear relax users,
> > > I am trying to run relax using spheroid model. I am a bit confused how to
> > set
> > > diffusion tensor. I tried r2r1_diffusion to get Diso and Dper and Dpar,
> > Theta
> > > and Phi, then input these values into the script
> > (sample_script/model-free.py)
> > > and in this script "diffusion_tensor.init(name, (2e-8, 1.3, 60, 290),
> > > param_types=0, spheroid_type='prolate', fixed=1)" I guess the values 
> should
> > be
> > > tm and Da, theta and Phi (is the second value 1.3 Da? seems not right 
> cause
> > I
> > > got 5e7 for this)
> >
> > It looks to me like this line in model-free.py was not properly updated
> > when the diffusion_tensor.init() function was changed in relax a few
> > versions back. The line should read:
> >
> > diffusion_tensor.init(name, (2e-8, 1.3, 60, 290), param_types=2,
> > spheroid_type='prolate', fixed=1)
> >
> > ie. the value 1.3 is actually Dr (= Dpar/Dper), not Da
> >
> > Sorry for the confusion.
> >
> > The full set of possible parameter sets and their corresponding
> > param_types values are described in the documenation for this function
> > (do help(diffusion_tensor.init) at the relax prompt, or look at the
> > manual)
> >
> > >  what do the flages mean? I tried to change param_types=0 to
> > > 1, relax will require me to change tm to Diso. Again if I change fixed 
> from
> > 1
> > > to 0, relax will encounte problems with GridsearchSize too big error and
> > stop.
> > > How should I set the diffusion tensor?Does it means if fixed=1, the
> > diffusion
> > > tensor I input should be fixed? But I checked results that the Theta and
> > Phi
> > > still varied.
> >
> > The fixed flag simply defines whether or not the diffusion tensor can be
> > varied during optimisation. This can be changed later using the fix()
> > command. Grid searches will fail unless the diffusion tensor is fixed
> > due to the dimentionality of the search space.
> >
> > Chris
> >
> >
> > >
> > > I would be very much appreciate any suggestions!
> > >
> > > Hongyan
> > >
> > > Dr. Hongyan Li
> > > Department of Chemistry
> > > The University of Hong Kong
> > > Pokfulam Road
> > > Hong Kong
> > >
> > >
> > > _______________________________________________
> > > relax (http://nmr-relax.com)
> > >
> > > This is the relax-users mailing list
> > > relax-users@xxxxxxx
> > >
> > > To unsubscribe from this list, get a password
> > > reminder, or change your subscription options,
> > > visit the list information page at
> > > https://mail.gna.org/listinfo/relax-users
> > >
> >
>
>
> Dr. Hongyan Li
> Department of Chemistry
> The University of Hong Kong
> Pokfulam Road
> Hong Kong
>
>
> _______________________________________________
> relax (http://nmr-relax.com)
>
> This is the relax-users mailing list
> relax-users@xxxxxxx
>
> To unsubscribe from this list, get a password
> reminder, or change your subscription options,
> visit the list information page at
> https://mail.gna.org/listinfo/relax-users