Biology Reference
In-Depth Information
to have been used in an evolved organism, but that might be a possible unexplored area of
phenotype space that could, in principle, turn up in a phylum if the history of the Earth were
run again. This type of exploration is similar to hypothesis testing but is much less formal
because, with a computer, it can explore large numbers of entirely random patterns of
'wiring'. Indeed, if a particular goal can be specified, and some scoring pattern for closeness
of approach to that goal can be devised, a connection of components suitable for that goal
could be found automatically using a genetic algorithm. In genetic algorithms, a computer
begins by testing how well each of a set of randomly constructed arrangements approaches
the desired goal. The winners are then used to generate another set that consists of the
winners themselves and random variations of them. In some systems, elements of the
winners can also be interchanged, in what is effectively a simulation of sexual reproduction.
The new sets of arrangements are then tested, the winners chosen and variations made, and
so on through generations of mutation and selection until the goal is achieved. If mimicry of
a real morphogenetic event is the aim, the winning arrangement of parts can be used to guide
the formation of a hypothesis about how the real system works, and the hypothesis can be
tested at the bench. The hypothesis will very often be wrong, for two possible reasons. The
first possibility is that the system really could work this way, but happened to evolve to
work some other way: this can be tested by building the computer-identified system in
a real cell and seeing if this generates the same type of morphogenetic event (this approach
is described in more detail in Chapter 27). The second possibility is that an unfeasible hypoth-
esis was generated because the assumptions that were implicit in the model were wrong:
'garbage in, garbage out' as the computer scientists' saying goes. This finding is valuable,
because it alerts a researcher that his or her understanding of some aspect of cell biology
must have been incorrect. Knowing some aspect of cell biological understanding is inad-
equate is a valuable spur to progress in finding out how that aspect of a cell really does work.
BROAD STRATEGIES FOR MODELLING:
MA
THEMATICAL VERSUS SYNTHETIC BIOLOGIC
AL
There are two main strategies for modeling that are sometimes used together. The 'trad-
itional' one is to use a physical or mathematical system in which every aspect of the model
is known and can be monitored or altered. Various ways of constructing such models, and
examples of their use making a significant contribution to understanding, are presented in
Chapter 26. The main advantages of such models are that they are relatively quick and inex-
pensive to construct and they allow the system represented to be truly isolated from all bio-
logical interference. This is also their disadvantage. Computational approaches allow
a modelled system to appear to drive morphogenesis in a robust and reliable way when,
in fact, it might be completely unreliable in the environment of a real living cell. Sophisticated
computer models can assess robustness against known sources of interference, for example
thermal ('kT') noise and random movement of cells, but they cannot test against interactions
with other cellular systems that nobody suspects exist: it is the problem of unknown
unknowns.
An alternative type of modelling, which has roots going back into the twentieth century
but is now a rapidly growing field, is the use of synthetic biological techniques to introduce
