>Gregory King (GKING at ARSERRC.GOV) wrote: : Greetings,
>> I'm looking for software (or an explanation of an efficient
> algorithm) that will superimpose molecule A onto molecule B with
> a combination of translations and rotations, and using a minimal
> rms deviation as the convergence criterion. There is a known
> one-to-one correspondence between the relevant atoms in both
> molecules (I'll be comparing peptides, and using the backbone atoms
> as the relevant ones).
To which wellsc at river.it.gvsu.edu responded:
> I would point out that an iterative method is not necessary, and in
> fact is wasteful of computing time. A method which has been known in
> robotics for at least ten years uses the singular value decomposition
> of a 3 by 3 matrix to yield a closed form (rather than iterative)
> procedure.
I thought the standard way to do rmsd best-fit alignment of two structures
with 1-1 assignment, as in Kabsch, Acta Cryst. (1978) A34, 827-828,
and that's what I used in our visualization program, VMD. It isn't
iterative and works fine as far as I can see.
Andrew
dalke at ks.uiuc.edu
