On 28 Aug 2000, at 9:49, Petr Kuzmic wrote:
>[...]
> I also understand the need to move away from mechanisms in those cases
> where we do not know what they are. For example, we've all see enzyme
> reactions with a complex substrate-velocity dependence similar to this
> (sigmoidal increase to a maximum, followed by a decrease to an
> asymptote):
> [...]
> This behavior could have many different mechanistic interpretations, so
> why not just describe it by some rational polynomial,
>> V = (a + b[S] + c[S]^2 + ...) / (d + e[S] + f[S]^2 + ...),
>> determine the coefficients a, b, c, ... from the available data, and use
> the polynomial model without ever even trying to ascribe physical
> meaning to a, b, c, etc. But what are the limitations of this empirical
> (as opposed to mechanistic) model building? Isn't it true that the
> empirical model so built is only applicable to the exact conditions that
> were used to construct it?
Well, if the conditions are sufficiently general, that should be no
problem. In other words, if the range of S-concentrations that
were used to fit the above polynomial is wider than the S-
concentrations calculated by a larger model in which you use
this rate eqn, you should be OK. Of course, if [S] moves either
higher or lower, we have a problem with emprical fitting, as the
best fit over a range of data often has the worst predictive power
beyond the data range.
However, "mechanistic" rate equations suffer from similar
shortcomings as emprically fitted ones. Km's, Ki's, Hill
coefficients, etc., fitted on experimental enzyme kinetic data often
depend on temperature, ionic strength, pH, etc. It is very seldom
that these conditions are exactly physiological. Therefore, if you
assemble those kinetic data into a model to simulate the
behaviour of a metabolic pathway in vivo, you are equally
restricted whether you use mechanistic kinetic parameters or
empirically fitted ones. In both cases, the parameters may not
truly reflect the kinetic behaviour of the enzyme beyond the
experimental conditions that were used to determine them.
As you point out, the only advantage is that mechanistic
parameters have a physical interpretation (affinity, etc.). But if we
want to model larger pathways, this may not be of such
importance.
Johann
--
Johann M. Rohwer
Dept. of Biochemistry, Univ. of Stellenbosch
Private Bag X1, 7602 Matieland
SOUTH AFRICA
Tel.: +27-21-8085843
Fax: +27-21-8085863
E-mail: jr at maties.sun.ac.za
WWW: http://www.sun.ac.za/biochem/
--
---