On Thu, Jul 29, 2010, InSuk Joung wrote:
>
> Recently I posted a problem about the calculation of dihedral energy using
> NAB to AMBER mailing list. I think I found the source of the problem
>
> At the 458th line of eff.c:
>
> uu = (aa * bb) - (ab * ab);
> vv = (bb * cc) - (bc * bc);
> uv = (ab * bc) - (ac * bb);
> den = 1.0 / sqrt(uu * vv);
> co = uv * den;
>
> When 'uu' or 'vv' is 0. 'den' goes to infinity, which causes a problem. I
> am not sure if this way to calculate the dihedral angle is computationally
> efficient but I believe dihedral angles can be calculated using atan2
> without the numerical problem. Can anybody suggest a detour for this?
First, what is the geometry of the four atoms you are using here? Is either
the 1-2-3 or the 2-3-4 angle 180 degrees? (Dihedral torsions are not well
defined in such cases).
Second, when uu or vv are zero, is uv also zero?
....dac
_______________________________________________
AMBER-Developers mailing list
AMBER-Developers.ambermd.org
http://lists.ambermd.org/mailman/listinfo/amber-developers
Received on Thu Jul 29 2010 - 19:30:06 PDT
