Re: Bug in structure.create diff tensor pdb -- March 13, 2009

Edward d'Auvergne wrote:

On Thu, Mar 12, 2009 at 7:04 PM, Sébastien Morin
<sebastien.morin.1@xxxxxxxxx> wrote:

Edward d'Auvergne wrote:

On Thu, Mar 12, 2009 at 6:05 PM, Sébastien Morin
<sebastien.morin.1@xxxxxxxxx> wrote:

Hi,

In an attempt to find the problem causing this bug, I dug into the code
and found something which could be erroneous... This alone does not fix
the bug, but may be part of the fix (or not)...

In function math_fns/direction_cosine.calc_ellipsoid_di(), there could
be an error at line 229:

   diff_data.dz[1] =  data.sin_b * data.sin_g

If I am right (?), this part of the code is related to page 193 of
Edward  d'Auvergne's thesis:

   c11  =  diff_data.dx[0]
   c21  =  diff_data.dx[1]
   c31  =  diff_data.dx[2]

   c12  =  diff_data.dy[0]
   c22  =  diff_data.dy[1]
   c32  =  diff_data.dy[2]

   c13  =  diff_data.dz[0]
   c23  =  diff_data.dz[1]  -->  FALSE !!!!
   c33  =  diff_data.dz[2]

Indeed,
   c23 = sin(beta) * cos(gamma),
whereas
   diff_data.dz[1] =  data.sin_b * data.sin_g


Thus, either me, the code or Edward's thesis are erroneous here...

Am I right ?

These unit vectors are the columns of the reverse matrix, or the
transpose, or simply rows of the normal matrix.  This depends if you
are defining the rotation from being from the diffusion tensor frame
to the lab frame or the reverse - both are related by the transpose.
Hence this value is c32 which in my thesis is sin(alpha)*sin(beta).
Does this answer your question and solve the issue?  There could be
bugs elsewhere in here though because of this.

Ok... I am far from an expert in trigonometry...  :p

If I understand well what you wrote, there is no error in the code or
the thesis ? Or should something be corrected ?


Sorry, I had a much closer look and you are correct - there is an
error there.  The problem is a typo in equation 6.61 (page 193) of my
thesis, c23 should be sin(beta) * sin(gamma).  I double checked using
wikipedia and tracked it back to the thesis.  Glad that it's in the
thesis and not the code!!!

Cheers,

Edward


P.S.  It would be useful if you could confirm that the problem is
indeed in the thesis and not the code ;)

I agree.

The error in not in relax, but in the thesis... I verified this on the
Mathematica website at:
    http://mathworld.wolfram.com/EulerAngles.html

where equations 39-44 are equivalent of equation 6.61 in your thesis
(page 193, c_23 -> sigma_23). From Mathematica:

    sigma_23 = sin PSI * sin THETA

thus

    c_23 = sin GAMMA * sin BETA

is what should be found in the thesis (instead of c_23 = cos GAMMA * sin
BETA).

Note that everything is fine in relax code.  :)

Ok, I'll correct my printed copy of your masterpiece..!

Regards,


Séb  :)

-- 
Sébastien Morin
PhD Student
S. Gagné NMR Laboratory
Université Laval & PROTEO
Québec, Canada

Re: Bug in structure.create diff tensor pdb

Header

Content

Related Messages