Biology Reference
In-Depth Information
middle-out (e.g. Brenner, 2001; Noble, 2003). While the true understanding of
complex living systems and/or their subsystems will likely involve the judicious
and iterative blending of each, it is convenient to use this distinction as a means
of discriminating the necessary methodologies.
Analytical or top-down systems biology tends to start from the system as
a whole. In a way it comes from the direction of holism and moves towards
molecular mechanism. Either from empirical relations between genome-wide
patterns of gene expression, or by calculating properties of genome-wide net-
works, it induces or proposes the occurrence of more general principles, such
as the feature that metabolic networks correspond to small world, scale-free
networks (Barabási & Oltvai, 2004; Wagner & Fell, 2001) and that genetic net-
works abound in certain regulatory motifs (Itzkovitz & Alon, 2005; Milo et al.,
2002; Yeger-Lotem et al., 2004). These views may then be tested.
In the leaner, 'Synthetic' or bottom-up branch of systems biology, one typi-
cally starts with a qualitative ('structural') and often simple model of molecules
interacting with each other in networks, then seeks to determine what system
properties might emerge from the nonlinear interactions. By then parameterizing
the equations that describe these interactions and inserting parameter values that
correspond to actual subsystems, more or less realistic predictions of system
properties are achieved. When the predictions are accurate, the proposed mech-
anisms of emergence of the functional properties are considered to have become
more likely. This method is reductionist in that it prefers to deal with simple
parts of the true system but not so simple as to lose important aspects of the
interactions and the emergence of interesting functional properties. 'Bottom-up'
methods start with purified entities (e.g. proteins) that allow the measurement of
the parameters, while 'top-down' methods seek to infer their values via 'reverse
engineering' of the parameters values through fitting of the calculated system
behavior to experimentally observed system behaviour.
4.3.3. The bottom-up approach to systems biology
Our own prejudices - given a historical focus more on metabolic than signalling
systems (Kell et al., 1989; Kell & Westerhoff, 1986; Mendes et al., 1996;
Pritchard & Kell, 2002; Raamsdonk et al., 2001; Teusink et al., 2000; Westerhoff
& Kell, 1987; Westerhoff & Kell, 1988; Westerhoff & Kell, 1996; Westerhoff
et al., 1991), and on unicellular organisms rather than the more obviously (cf.
Davey & Kell, 1996; Kell et al., 1991) differentiated 'higher' organisms - leads
us to concentrate more on the 'bottom-up' approach (Fig. 5), embodied in the
'silicon-cell' concept (Westerhoff, 2001): if we can measure all of the 'local'
properties of individual players in a complex system, including their interactions,
we can bolt the system together and whatever new properties may emerge will
indeed emerge and produce the 'whole system' properties that can indeed be
compared with those of the intact system. The apotheosis of this approach to
