Biomedical Engineering Reference
In-Depth Information
2. Modeling helps identify essential components, i.e., the major players of a
given system, and to filter out redundant (i.e., nonessential) elements, which
may represent evolutionary "debris."
3. Modeling facilitates rejection of false hypotheses and enables a more
precise understanding of the non-intuitive behavior of a system.
4. A good model can accurately predict the future state of the system in the
presence/absence of a perturbation.
5. We can easily knock-in or knock-out components from the system and
study their upstream/downstream effects. A model is an inexpensive alternative
to wet lab tests.
3.2. Limitations of Modeling
1. Modeling may result in duplication of experimental results.
2. The incompleteness of knowledge may result in limited predictive power.
3. Models are often constructed to answer very specific questions, without
considering the big picture. Though somewhat unavoidable, the model must
grow with time.
4. Even a good model may sometimes yield incorrect predictions. For ex-
ample, the model of T7 phage hinted at a connection between genomic rear-
rangement and its growth rate, a prediction that proved incorrect (7).
3.3. Mathematical Basis
Translating biochemistry into mathematics is what eventually drives com-
puter simulation, but this is not always a straightforward process unless the data
are clean. Modeling with differential equations enables extrapolation to future
states. Generally, analytical models are built with ordinary differential equations
and/or stochastic equations, the accuracy of which depends on the assumed pa-
rameters, rate laws, and concentrations (see Table 1). In order to find missing
parameters, various mathematical approaches are used: for example, simulated
annealing, genetic algorithms.
Advantages
: Precision, does not permit vague/fuzzy statements, combines
explanation with predictions, connects various levels of biological organization
Limitations
: nonlinearity (exhibited by biological systems) is very hard to
solve analytically.
3.4. Traits of a Good Model
An ideal model is experimentally validated, analyzable, and open for ma-
nipulation and optimization. The quality of a model is directly proportional to
the quality of the data. Good data are taken under uniform experimental condi-
tions, are based on time-series measurements, and are obtained by standard ex-
perimental protocols. The rule of the thumb includes (a) keep it as simple as
possible and as complex as necessary, (b) strive for realistic goals, (c) do not
