From: Borries Demeler <demeler@bioc09.v19.uthscsa.edu>
  To  : John Correia <jcorreia@biochem.umsmed.edu>
  Date: Tue, 16 Nov 1999 19:27:32 -0600 (CST)

Re: s coeff and shape

> 2nd - this requires looking at old lit - but s(c) is typically linear over =
> at least one order of magnitude.  In those days a wide range of conc were =
> studied typically by schlieren & thus over the range 2 to 20 or 30 mg/ml.  =
> The Ks magnitudes observed obviously are derived from this range.

In my experience the ks is linear only for certain samples in this range
and may become nonlinear much earlier for *certain* samples. Unfortunately,
I don't have enough information to predict when this happens. Certainly for
DNA we have seen nonlinear ks for concentrations even lower than 1 mg/ml,
especially for long, linear molecules in low salt. This may be extreme,
but similar conditions could potentially apply for proteins.

> 3rd - If a protein exhibits nonideal behavior by a vanHolde-Weischet plot, =
> than I think the experimental questions are:
> 
>   1)  how asymmetric or charged must it be to demonstrate significant =
> S(C)) at 1 mg/ml.   Or when is a protein like nucleic acid.   Any good =
> examples out there like monomeric myosin which I think is not all that =
> nonideal when compared to DNA???

Good question - if anybody has any example data on this, I would like
to take a look at it with vHW.

> Finally, we have observed nonideal boundaries by vanHolde-Weischet =
> analysis that is artifactual due to two causes.  1)  loss of material from =
> the plateau due to aggregates.  Bo has documented this in his BJ paper I =
> believe.  

I would like to add stray light and poor intensity or misaligned optics,
they can result in a sloping baseline or nonlinear absorbance response
which can also give artifactual data (this causes problems with all
analysis methods, not just vHW. In my finite element version, and I believe
in John Philo's Svedberg program this can be adjusted as a separate 
parameter).

> 2) collecting data at less than maximum point density - you must =
> use 0.001 cm to get a good vanHolde-Weischet plot.  I do not understand =
> the origin but using a spacing of 0.003 or more gives false nonideality =
> (in our hands - anyone else??).  With the XLA this often means running one =
> sample at a time, especially if you wish to do significant signal =
> averaging.

I have not seen this, and we measure at both .003 and .001 cm settings
(the real data is more like 0.0035 and 0.0015 on the average due to
non-constant radial spacings). If you have such 0.001 and 0.003 cm data,
I would appreciate it if you could zip it up and let me have a look at it.

Regards, -Borries