1
Introduction
2
The field of food microbiology is a very broad one, encompassing the study of
3
microorganisms which have both beneficial and deleterious effects on the quality and
4
safety of raw and processed foods. The primary tool of microbiologists is the ability to
5
identify and quantitate food-borne microorganisms; however, the inherent inaccuracies
6
in enumeration processess, and the natural variation found in all bacterial populations
7
complicate the microbiologist’s job. Accumulating sufficient data on the behaviour of
8
microorganisms in foods requires an extensive amount of work, and is costly. In
9
addition, while data alone can describe the response of a microorganisms in food, they
10
provides little insight into the relationship between physiological processes and growth
11
or survival. One way in which this link can be made is through the use of mathematical
12
models.
13
14
In its simplest form, a mathematical model is a simple mathematical description of a
15
process, and models have been used extensively in all scientific disciplines. They were
16
first used in food microbiology in the early 20th century to describe the inactivation
17
kinetics of food-borne pathogens during thermal processing of foods. Since then, with
18
the advent of personal computers and more powerful statistical software packages, the
19
use of modelling in food microbiology has grown to the point of being recognized as a
20
distinct discipline of food microbiolgy, termed “predictive microbiology”. This concept
21
was introduced and extensively discussed (with particular reference to growth of food-
22
borne pathogens) by McMeekin and his colleagues at the University of Tasmania [3].
23
24
Predictive microbiology has application in both microbial safety and quality of foods;
25
indeed, early development of the concept was based on seafood spoilage. Extensive
26
research in recent years has shown, however, that the most important application of
27
predictive microbiology is in food safety. Mathematical models for microbial growth and
28
survival are now sufficiently detailed and accurate to make an important contribution to
29
the ability of scientists and regulators to make reasonable predictions of the relative risk
30
posed by a particular food or food process. It has been argued, however, that predictive
31
microbiology is a misnomer; predictive microbiology does not actually make predictions
32
at all. To examine this further, we need to first look at some definitions of modelling.
33
34
In a general sense, a “model” is a simplification of a system using a combination of
35
descriptions, mathematical functions or equations, and specific starting conditions.
36
There are two general classes of models: descriptive and explanatory, the latter being
37
composed of analytical and numerical models. Descriptive (or observational, empirical,
38
“black box” or inductive) models are data-driven; approaches such as polynomial
39
functions, artificial neural nets, and principle component analysis are used to classify
40
the data. True predictions with this class of models are difficult to make, since they
41
cannot be extrapolated beyond the data used to build the model. In spite of this,
42
descriptive models are used extensively with considerable success in predictive
43
microbiology.
44
45
Explanatory (or mechanistic, deductive, or “white box”) models aim to relate the given
46
data to fundimental scientific principles, or at least to measureable physiological
47
processes. Many predictive microbiology models have parameters which are related to
48
observed phenomenon, and are considered mechanistic. The common form of these
49
are analytical models, explicit equations which can be fit to data. To be truly
50
mechanistic, however, a model should raise new questions and hypotheses which can
51
be tested. This is not always easy with explicit functions, since it is difficult to extend
52
them to dynamic situations, or to add additional steps to the model.
53
54
Numerical approaches are designed specifically to allow further development of the
55
model. These models are hierarchial, containing submodels at least one level deeper
56
than the response being described. They have been extensively developed for complex
57
ecological systems, but have not been applied to any great extent in predictive
58
microbiology. Numerical models have been difficult for non-mathematicians to apply,
59
due to the need for extensive programming skills. New software platforms and
60
concepts such object-oriented programming have provided new tools for the
61
microbiologist to extend traditional modelling approaches.
62
63
The models discussed so far (and the majority of existing predictive microbiology
64
models) have all been deterministic. In a deterministic model, knowledge of the starting
65
conditions combined with a mathematical function describing the behaviour of the
66
system over time is sufficient to predict the state of the system at any point in time.
67
Bacteria, however, are not so cooperative. While it may be possible to define a
68
function, the starting conditions are less clear, particularly when dealing with individual
69
bacterial cells. Models which recognize and account for uncertainty or variability in an
70
experimental system are called stochastic or probabilistic models. Probability models
71
have been used extensively in the past to predict the probability of germination of
72
pathogenic spore forming bacteria. Recently, the behaviour of individual bacteria has
73
been likened to that of atoms, in a concept referred to as ‘quantal microbiology’
74
analgous in some ways to quantum mechanics [2]. In addition, the effect of
75
environmental stress on microorganisms leads to interpopulation diversity, where
76
individual cells may be phenotypically but not genetically different from each other [1].
77
Recent advances in the use of software which allows the development of probability
78
models has helped to make these approaches more accessible, and has provided more
79
support for the development of risk assessment procedures.
80
81
The excellent monograph by McMeekin and colleagues published in 1993 [3] has set
82
the stage for an explosive increase in predictive microbiology research in the last
83
decade of the 20th century. As we move into the next century (and millenium), there is a
84
need for a definitive work on the subject, above and beyond the many comprehensive
85
and stimulating reviews which have appeared. This book can serve many purposes.
86
Firstly, we believe it will be a primer for many who are not familiar with the field.
87
Chapters 1 (Experimental Design & Data Collection), 2 (Primary Models), and 3
88
(Secondary Models) are designed in part to give the uninitiated sufficient information to
89
start developing their own models. Other chapters address more complex issues such
90
as the difficulties in fitting models (Chapter 4: Model Fitting and Uncertainty), and the
91
relevance of models to the real world (Chapter 5: Challenge of Food and the
92
Environment). Extensive applications of predictive microbiology are covered in
93
Chapters 7 (Software Programs to Increase the Utility of Predictive Microbiology
94
Information) and 8 (“New chapter name”). The important contribution made by
95
predictive microbiology to quantitative risk assessment is describe in Chapter 8 (Risk
96
Assessment), and the further complication of individual cell behaviour and inter-cell
97
variability are addressed in Chapter 9 (Modelling the History Effect on Microbial Growth
98
and Survival). The future of predictive microbiology is the subject of Chapters 10
99
(Models: The Next Generation) and 11 (Conclusions).
100
101
We have attempted to cover the basics of predictive microbiology, as well as the more
102
up-to-date and challenging aspects of the field. There are extensive references to
103
earlier work, as well as recent publications. It is hoped that this book will reflect the
104
extensive research which has gone into placing predictive microbiology at the forefront
105
of food microbiology, and will provide stimulating discussion which will chart our way
106
forward.
107
108
References:
109
110
111
112
1. Booth, I.R. Stress and the single cell: Intrapopulation diversity is a mechanism to
ensure survival upon exposure to stress. Int.J.Food Microbiol. 78: 19-30 (2002).
113
114
2. Bridson, E.Y., & Gould, G.W. Quantal microbiology. Lett.Appl.Microbiol. 30: 95-98
(2000).
115
116
3. McMeekin, T.A., J.N. Olley, T. Ross, and D.A. Ratkowsky. 1993. Predictive
microbiology: Theory and application. John Wiley & Sons Inc., New York.
117