Professional Documents
Culture Documents
Articles
This paper has arisen as a result of teaching Models in Biology to undergraduates of Bioengineering at the
Gediminas Technical University of Vilnius. The aim is to teach the students to use a fresh approach to the
problems they are familiar with, to come up with an articulate verbal model after a mental effort, to express
it in rigorous mathematical terms, to solve (with the aid of computers) corresponding equations, and finally,
to analyze and interpret experimental data in terms of their (mathematical) models. Investigation of
microbial growth provides excellent possibilities to combine laboratory exercises, mathematical modeling,
and model-based data analysis. Application of mathematics in this field proved to be very fruitful in getting
deeper insight into the processes of microbial growth. The step-by-step modeling resulted in an extended
model of the growth covering conventional “lag,” “exponential,” and “stationary” phases. In contrast to
known models (differential equations that can be solved only numerically), the present model is expressed
symbolically as a finite combination of elementary functions. The approach can be applied in other areas
of modern biology, such as dynamics of various cellular processes, enzyme and receptor kinetics, and
others.
Keywords: Verhulst equation, logistic equation, growth rate, generation time, lag time.
Performing laboratory exercises, students learn various material resources or skills or special knowledge of math-
techniques and acquire practical skills, whereas while ematics) can be easily conducted, whereas experience
modeling, they apply all their experience, above all, their gained in these exercises and modeling will be very useful
knowledge of mathematics. Students of biochemistry and in other areas. The appreciation of role and importance of
molecular biology, among others, are (or at least are sup- models in science and engineering evolves as a result of
posed to be) computer-literate nowadays. They are sup- this activity.
posed to have basic mathematical knowledge as well. Still, At the Gediminas Technical University of Vilnius, under-
some students seem to have “an aversion to math, ’Math- graduates of bioengineering have a four-credit course of
ematophobia,’” having “little appreciation for the impor- Models in Biology concurrently with biochemistry, micro-
tance of models in scientific progress” [1]. Mathematics biology, and other courses with laboratories. As a result of
does not seem to be properly appreciated by teachers (or teaching (and learning), they are expected to use a fresh
textbook writers), either. In textbooks, mathematical equa- approach to the problems they are familiar with, to come
tions, or “formulae,” if presented at all, often appear to be up with an articulate verbal model after a mental effort, to
just a mandatory tribute to modernity. It is more so in express it in rigorous mathematical terms, to solve (with
microbiology textbooks (e.g. see Refs. [2– 4]), where pres- the aid of computers) corresponding equations, and, fi-
entation of microbial growth is most primitive, without any nally, to analyze and interpret experimental data in terms
reference to the growth equation (known from the mid- of their (mathematical) models. Students are requested to
19th century as the Verhulst or logistic model); neither submit their work in electronic form rather than hard copy.
“Verhulst” nor “logistic” can be found in the index. (The The students are encouraged to take part in annual scien-
“lack of interdisciplinary communication” [5] is sad tific conferences (e.g. see Refs. [8 and 9]). One of the
enough; it is more so in microbiology, the lack being intra- present writers (R.I.) is a student.
disciplinary. The growth equations seem to be ignored in We consider it pedagogically relevant to present step-
the textbooks, whereas being considered in scientific jour- by-step modeling of microbial growth and model-based
nals (e.g. see Refs. [6 and 7])). On the other hand, research analysis of experimental data in the main text rather than
in the population growth provides excellent possibilities for putting the modeling in an appendix. Our endeavor in
both laboratory exercises and qualitative and quantitative modeling has been rewarded by what amounts to a little
data analysis; such studies (not being highly demanding of discovery. The extended model appears to be new (it has
never been published before) and rather simple. It can be
‡ To whom correspondence should be addressed: E-mail: extended further on. The way of getting such a model is of
alfonsas@reda.vtu.lt. interest far beyond the area of microbiology education.
This paper is available on line at http://www.bambed.org 417
418 BAMBED, Vol. 34, No. 6, pp. 417–422, 2006
MATHEMATICAL MODELING
It is clear that at any moment, t, (absolute) rate, a, of
growth of a population size (number of cells, total mass,
(optical) density, etc.), A, is proportional to the population
size, i.e.
a ⫽ ␣A (Eq. 1)
the coefficient of proportionality, ␣, being relative rate of
population growth (or specific growth rate). The statement
above (a verbal model of growth, as a matter of fact) can
be expressed mathematically in the form of the differential
equation,
dA
⫽ ␣A (Eq. 2)
dt
The specific growth rate depends on the environmental
conditions, and it may be considered that
␣ ⫽ ␣0 共␣0 ⫽ const兲 (Eq. 3)
as long as the conditions remain unchanged. With the
growth of population size (density), the conditions may
change because of the growth; first of all, the organisms may
have to compete for available resources. Let the initial sup-
ply of the resources equal to 1. This supply may be sup-
posed to be depleted proportionally to the population size;
the availability of resources, therefore, will remain 1 ⫺ A/B,
and, as a consequence, the relative rate of growth will be,
冉 冊
␣ ⫽ ␣0 1 ⫺
A
B
(Eq. 4)
siderably from further experimental data (a larger part of the right-hand part of Equation 4, we obtain immediately a
the data, as a matter of fact; Fig. 1A). However, before model that is equivalent to the Verhulst or logistic equation
discarding the model, let us note that according to this (the equivalence can be verified by Maple as well; Fig. 2B).
model, at the moment of T0 (which has yet to be found), the We present the equation in a slightly different form
population doubles, i.e.
A0e␣0t
A0e␣0T0 ⫽ 2A0 (Eq. 6) A⫽ (Eq. 9)
A0
1 ⫹ 共e␣0t ⫺ 1兲
from where (taking the natural logarithm of both sides of B
the equation)
which is more convenient for analysis and comparing with
ln 2 other models. (Other models are presented in a similar
T0 ⫽ (Eq. 7)
␣0 way, differing from the immediate Maple output. Although
appearing rather complicated, they can be visualized di-
Parameter T0 is called the generation time. It follows from rectly in a Maple worksheet as graphs of population size
Equation 6 (by dividing both parts of that equation by A0 versus time or copied and pasted into Excel worksheet.)
and raising them to power n) that Note that the numerator of the above model is the right-
hand part of Equation 5. It can be seen that for t 3 ⬁ A 3
en␣0T0 ⫽ 2n (Eq. 8)
B, thus the meaning of parameter B; it is the asymptotic
That means that whereas the model (Equation 5) is con- value of population size. It would be reasonable, therefore,
tinuous, the doubling in the number of microorganisms as to refer to Equation 9 as a model of limited growth (MLG)
a result of binary fission at discrete time intervals, T0, is in contrast to Equation 5, which might be designated a
essentially a discrete process making a geometric pro- model of unrestricted growth (MUG). For B ⬎⬎ A0, MLG is
gression. Analysis of microbial growth provides, therefore, close to MUG, and for B 3 ⬁, it is reduced to MUG (Figs.
an excellent opportunity to demonstrate once more that 1 and 2C).
(discrete) geometric progression can be extended by a Once a model has been accepted as “good enough” to
more general (continuous) exponential function. approximate experimental data, all the information con-
Students may need to be reminded of how to solve the tained in the data should be considered to be contained (in
above equation or just shown how to use popular com- a generalized mathematical form) in the model, there being
puter programs such as Mathcad or Maple [10]. Without no need ever to come back to the data; everything of
entering into a detailed discussion on how to use Maple in interest can be derived from the model. The latter can be
general or its computing possibilities, students are pre- used to model the growth under conditions (technically)
sented with a working example (Fig. 2) of solving the first inconvenient for direct experimental investigation. In Fig.
order differential equation, the right-hand part of which 3, the right-hand curves correspond to the inoculum size
can be substituted by any reasonable expression. Entering about a thousand times lower than that used in the above
420 BAMBED, Vol. 34, No. 6, pp. 417–422, 2006
(the right-hand part of Equation 9); this rate divided by the
population size (the right-hand part of the same equation)
yields the time course of the relative (or specific) growth rate,
共B ⫺ A0兲e⫺␣0t
␣ ⫽ ␣0 (Eq. 10)
共B ⫺ A0兲e⫺␣0t ⫹ A0
as shown in Fig. 1, curve b⬘. It can be seen that the relative
rate declines over time, the initial rate (at t ⫽ 0) being
冉
␣initial ⫽ ␣0 1 ⫺
B冊
A0
(Eq. 11)
再
growth, and those phases by implication possess their
d⫹g⫽1 specific characteristics or parameters. Indeed, the stu-
dg (Eq. 12) dents often ask about the duration of the phases, espe-
⫽ d
dt cially about that of the lag phase. The lag is of great
Solution of this system with respect to g taking into ac- importance in food microbiology; in the papers of the field,
count the initial condition (let g(0) ⫽ G (0 ⱕ G ⱕ 1), i.e. at it is often denoted as L or . However, as Buchanan et al.
the moment of inoculation portion G of the population is [7] point out, “there is currently no generally accepted
growing) results in definition for lag phase that is based on physiological
events occurring in the bacterial population.” Estimation of
g ⫽ 1 ⫺ 共1 ⫺ G兲e⫺t (Eq. 13) this parameter, as well as its incorporation into growth
equations, is bound, therefore, to serious difficulties. The
Now Equation 2 can be modified taking into account both reason of these difficulties, as Baty and Delignette-Muller
Equation 4 and Equation 13. [11] put it, “is the lack of physiological understanding of the
dA
dt
A
⫽ ␣0共1 ⫺ 共1 ⫺ G兲e⫺t兲 1 ⫺ A
B 冉 冊 (Eq. 14)
lag phenomenon. In actual fact, little knowledge is avail-
able concerning this physiological stage and only few au-
thors were able to put some biological information about
The solution of the above equation is into model equations. The second reason is related to the
⫺t共1⫺G兲␣0/
first one and comes from the fact that the actual definition
A0e␣0t⫹e of is either purely geometric or purely mathematic.” The
冉 冊
A⫽ (Eq. 15)
共1⫺G兲␣0/
A0 ␣0t ⫹ e⫺t共1 ⫺ G兲␣0/ very approach (the attempts to incorporate into model
e ⫹ e ⫺ e共1 ⫺ G兲␣0/ equations) implies tacitly the lag to be of “primary” or
B
“fundamental” order. Such a “hidden assumption” [5]
where A0, as above, is inoculation size comprised of the seems to be more a hindrance than a help in getting at the
cells both in dormant and in growing states. As can be essence of the processes that lead to the lag. The lag can
seen, for G ⫽ 1 (at the moment of inoculation, all the be redefined painlessly on the basis of Transition 1 (con-
population is in the growing state) or  ⬎⬎ ␣, Equation 15 sidered above) both in mathematical (probabilistic) and in
is reduced to Equation 9. Similarly, for B 3 ⬁, it is reduced geometrical terms.
to Equation 5. For G ⫽ 0 (at the moment of inoculation, all Indeed, the lag is caused, presumably, by the inability of
the population is completely in the dormant state), it microorganisms residing in state D (dormant) to take part
becomes in the growth of the population, the duration of the lag, ,
⫺t being determined by the duration of residence of the or-
A0e␣0t⫹e ␣0/
冉 冊
A⫽ (Eq. 16) ganisms in state D. The transition D 3 G, presumably, is a
A0 ␣0t⫹e⫺t␣0/ random process, the duration of residence of an individual
e ␣ 0 / ⫹ e ⫺ e ␣ 0 /
B organism in state D being a random value whose proba-
bility distribution can be found the following way.
As can be seen for  ⬎⬎ ␣0 (very high rate of transition from
The relative number of the organisms, d, residing in state
the dormant to the growing state), both Equation 15 and
D in the course of time is the solution of Equation 12 with
Equation 16 are reduced to Equation 9. Equation 16 was
respect to d taking into account the initial conditions. Let
fitted to the experimental data without any significant de-
d(0) ⫽ 1 (i.e. at the moment of inoculation, all the orga-
viation (Fig. 1, curve c and closed circles). Equations 15
nisms are in state D). Thus,
and 16, therefore, can be considered to approximate ex-
perimental data adequately. The extended model (EM), d ⫽ e ⫺ t (Eq. 18)
therefore, being more general than MLG, in addition to the
latter, also covers the “lag phase.” By analogy with Equa- This solution multiplied by  can be considered, therefore,
tion 9, the dynamics of the relative growth rate can be as density of probability distribution of the duration of
determined for EM as well, residence of the organisms in state D.
⫺t␣01/
The mean duration of residence of the organisms in this
共B ⫺ A0兲共1 ⫺ e⫺t兲e⫺␣0t⫺e state, therefore, is
␣ ⫽ ␣0 ⫺␣0t⫺e⫺t␣0/ (Eq. 17)
(B ⫺ A0)e ⫹ A0e⫺␣0/
冕
⬁