From: Julius.Clauwaert <clauwaer@uia.ua.ac.be>
  To  : rasmb@bbri.harvard.edu
  Date: Mon, 15 Dec 1997 13:59:30 +0100 (MET)

artractive molecules?

Dear  RASMB members.

We encountered already twice the same problem in similar conditions so we
believe there is some basic mechanism for our findings.
We have been measuring the molar mass of proteins using equilibrium
sedimentation: after reaching equilibrium we always run the centrifuge
at high speed to determine the base line.
Routinely we use the origin software from Beckman or the program that 
Les Holladay has sent us (many thanks Les). In both cases we get very 
similar results.
If we accept an ideal solution, the programs give us too high molar masses
and the fits are not good at all; but if we extrapolate the obtained molar 
masses to a concentration zero, we get a very reasonable molar mass Mo.
We were lucky to know the molar mass in both cases: once from the
combination of s and D and once from the primary structure.
It looks that we are working in both cases with not ideal solutions and 
that the molecules are attracting each other, so the second virial 
coefficient is negative. So if we introduce an non ideal solution in our
fitting programs, we indeed get a negative virial coefficient.
But for the different concentrations of the same protein, we get quite
different B values and subsequent varying molar masses. It looks, that we
need extreme good experimental data in order to get any reasonable result
from the non ideal fit.
We have tried also a floating base line (although we know it) or a
combination of a floating base line and a second virial coefficient: 
in these cases, we get  a better fit parameter but the M values do not
seem to agree and the Mo value is further away from the exact value 
(10 to 20%).
So what can we do?
Is our interpretation about the quality of experimental data in order to
use a non-ideal fit, correct ?
Could we introduce some constraints about the B value (e.g. set the range) 
in order to improve the agreement between the fitted results for the
different concentrations of the same protein.
We wish all RASMB colleagues a merry christmas and a happy new year.
T.Aerts, J.Clauwaert
University of Antwerp
Universiteitsplein 1
2610 Wilrijk.
Belgium