Understanding a complex system requires integration and collective analysis of data from many
levels of organisation. Predictive modelling of biochemical systems is particularly challenging
because of the nature of data being plagued by noise operating at each and every level. Inevitably
we have to decide whether we can reliably infer the structure and dynamics of biochemical systems
from present data. Here we approach this problem from many fronts by analysing the interplay
between deterministic and stochastic dynamics in a broad collection of biochemical models.
In a classical mathematical model we first illustrate how this interplay can be described in
surprisingly simple terms; we furthermore demonstrate the advantages of a statistical point of view
also for more complex systems. We then investigate strategies for the integrated analysis of models
characterised by different organisational levels, and trace the propagation of noise through such
systems. We use this approach to uncover, for the first time, the dynamics of metabolic adaptation
of a plant pathogen throughout its life cycle and discuss the ecological implications.
Finally, we investigate how reliably we can infer model parameters of biochemical models.
We develop a novel sensitivity/inferability analysis framework that is generally applicable to a
large fraction of current mathematical models of biochemical systems. By using this framework to
quantify the effect of parametric variation on system dynamics, we provide practical guidelines as
to when and why certain parameters are easily estimated while others are much harder to infer. We
highlight the limitations on parameter inference due to model structure and qualitative dynamical
behaviour, and identify candidate elements of control in biochemical pathways most likely of being
subjected to regulation.