Scatchard fit program for M

Nick Holford nholford at ccu1.aukuni.ac.nz
Wed Jul 28 18:14:48 EST 1993


rdtst+ at pitt.edu (Richard D Thorne) writes:

>Nick Holford, Dept Pharmacology & Clinical Pharmacology

>> ......A program called LIGAND that
>> does the right thing using non-linear regression and solving for
>> multiple ligands and binding sites was written in the late 60s Munson
>> and Rodbard at the NIH. It has been ported in various forms e.g. SCATFIT
>> in PL/I and BIND in QuickBasic (DOS). I cannot help you with a Mac
>> implementationtion but just the warning about straight line fitters
>> which are guaranteed to get the wrong answer except when there is no
>> noise in the data.

>  In my experience the output from Rodbard et al program and using the logistic
> equation is very slight.  Errors can be minimized by weighting.  The refs
> below may be of interest.  Richard

This seems to miss the point. Scatchard analysis means to me the use of
the Scatchard transformation which involves linear regression of bound
vs bound/unbound. Because this transformation alters the error structure
of the data most simple straight line fitting programs will assume
homoscedasticity and thus bias the parameter estimates of Bmax and Kd
derived from the slope and intercept. Logistic regression is performed
on the untransformed data and usually allows some kind of weighting
scheme to correct for heteroscedasticity. So beware of simple straight
line fitters that use bound vs bound/unbound.

-- 

Nick Holford, Dept Pharmacology & Clinical Pharmacology
University of Auckland, Auckland, New Zealand
INTERNET: n.holford at auckland.ac.nz




More information about the Bio-soft mailing list