adeboer at chem.vu.nl wrote:
> Hello,
>> I am an analytical chemist, combining enzymes with analytical
> instruments. Now I have a question.
> If I will obtain Km and Vmax out of a michealis-menten plot, which type
> of plot is the best to use, I mean: lineweaver-burk, eadie-hofstee,
> hanes-woolf, direct linear...?
Transformation of data also transforms the measurement errors. For
example in Lineweaver-Burk coordinates, error for the data with the
smallest concentration becomes largest, at the same time these data
points are shifted far to the right of the plot and influence the fit
params particularly strong. As a result, fitted params (in particular
Km) are off. Therefore, linearisation should be used only for
presentation, not for fitting. Lineweaver-Burk is used most often for
presentation and therefore easiest to interprete.
For actual data fitting non-linear regression to the untransformed data
should be used. There are basically 2 algorithms available for this
purpose: Marquard-Levenberg and Simplex. In their properties these are
complementary: Simplex gives a better fit and is less sensitive to the
choice of starting parameters, but can give error estimates for the
parameters only by bootstrapping. Marquard-Levenberg gives error
estimates directly, but does not work as well. Most computer packages
use Marquard-Levenberg, for example Gnuplot or Sigma-Plot.
