From: Borries Demeler <demeler@bioc09.v19.uthscsa.edu>
  To  : Reversible Associations in Structural and Molecular Biology <rasmb@alpha.bbri.org>
  Date: Tue, 16 May 2000 17:20:15 -0500 (CDT)

Re: who has a good fitting function with which to compute Mw(r,app)?

Dear Les,

If you *have* to smoothe the data (a la splines or related method)
I would suggest a Gaussian smoothing kernel. This is what I use for 
veloc. data when using van Holde Weischet. It takes pretty high degrees of
smoothing before you actually introduce "shape changes" in the curve.
In the Gaussian smoothing kernel you do a weighted fit where closer
neighbor points are weighted more heavily then further away neighbors.
The exact amount of weight is determined by a Gaussian normal curve.
The ends are a bit problematic, I solve that by reflecting the Gaussian
back onto itself. 

Another thing you could use would be a linear combination of many
exponentials to approximate the data in a general least squares fit. It's
very easy to compute and avoids all the problems with the splines. This
would be preferred in my opinion to the gaussian kernel smoothing for
equil data.  However, both methods are insensitive to data point spacings
and missing "in-between" points.

I can send you C code for the algorithm.

Hope that helps, -Borries