Appendix 1 Glossary of Terms
Amplification: a process in which the response of the system to
a stimulus increases over time via internal processes
Automatable : An algorithmic method that requires no user
intervention.
Compartment : a chamber (real or hypothetical) in which the
variable is homogenous throughout and assumed to be instantaneously
mixed, for example the concentration in a compartment.
Controllable state : A state that is changed when the input
variables are changed (e.g. the concentration in the body is changed by
dose).
Damping: a process in which the oscillations in a system,
caused by some stimulus are reduced, restricted or inhibited over
time.
Edge : a graphical representation of movement or an action. This
is usually shown as a line with a bar at the end or an arrow.
Empirical function : A user defined function to describe a set
of input-output data with no (or very few) assumption about the
underlying processes that generated the data.
Extrapolation : The prediction of an outcome or response by a
model under conditions or values of input variables other than those
(typically outside of the range) upon which the model was built.
Full-order model: a model that is its original (full) size,
e.g. an original QSP model
Input-output (I/O): pertaining to the input variables (e.g.
dose, dose time, route, etc) and the output variables represented by a
set of response variables (e.g. haemodynamic variables such as blood
pressure and heart rate etc)
Model dimension : The number of state variables & parameters in
a model.
Model-order reduction : a method to reduce a large model into a
simpler model
Model-order : A representation of the size of a model which is
related to the number of parameters.
Node : a graphical representation of a compartment in a model.
This is usually shown as a box or a circle (depending on preference)
Nonlinear system : A system whose outputs are not proportional
to inputs. In a nonlinear system, small changes in the inputs may lead
to substantial changes in the outputs.
Observable state : A state (node) that when changed results in a
change in the observed response (e.g. changes in the concentration of a
drug results in changes in blood pressure).
Pseudo-data : A set of input-output data simulated (e.g. from
full-order QSP model) to reflecting its behaviour and is used in
training a surrogate model.
Quasi-steady state approximation : An approach to approximating
equilibrium systems with different time scales by assuming that rapidly
equilibrating states reach equilibrium instantaneously.
Reduced-order model : a model that is smaller in size (number of
states & parameters) than the full-order model from which it was
derived.
State : a variable in a model whose value evolves through time.
This is the same as a concentration in a pharmacokinetic model. It
differs from a compartment as the compartment is the location
(hypothetical or real) of the concentration. In a pharmacokinetic
example, a state is the concentration of a particular drug in a
particular compartment. A state is normally shown graphically as a node.
Structural Identifiability : A property of a model whereby no
two (or more) sets of parameters give identical model predictions. A
model that is identifiable can be used for parameter estimation (and
vice versa).
System agnostic : The method applied to the system is not
affected by/dependent on the nature of the system.