You are on page 1of 15

SIMULATION AND MODELING

Mathematical models in biology examples


Example 2: Developing a model to describe bacteria growth.

Why bacteria?
One important dynamical process in biology is the growth of organisms and populations of
organisms. In some species, individuals grow larger over time. But in all species,
reproduction is imperative for survival. An important mark of a successful species is growth
in its population size, meaning the number of individuals in the population.

The population growth of bacteria is relatively simple, at least under carefully controlled
environments in the laboratory. For example, bacterial populations increase rapidly when
grown at low bacterial densities in abundant nutrient. The population increase is due to
binary fission, where single cells divide asexually into two cells, subsequently the two cells
divide to form four cells, and so on. The time required for a cell to mature and divide is
approximately the same for any two cells.
Discrete dynamic system modeling
Bacteria population growth can serve as a basis for a simple discrete
dynamical system model that introduces some key components of modeling
dynamical processes. In developing a discrete dynamical system model for
bacteria growth, we illustrate some of the steps in formulating a model and
analyzing its properties, with the goal of using the model to deepen our
insight into the underlying physical or biological process.
Vibrio natriegens, or V
Natriegens for short, is a marine bacterium Time (min) Population Density

commonly found in the mud around estuaries. One 0 0.022


can manipulate the growth rate of V. natriegens in a 16 0.036
laboratory by using simple manipulations of the 32 0.060
48 0.101
experimental conditions. Our goal will be to build 64 0.169
discrete dynamical system models describing its 80 0.266
population growth under a few experimental
conditions. Measurements of bacterial
density at pH 6.25.
The units of “Population Density”
Bacterial growth data from a V. natriegens are those of absorbance as
experiment are shown in the table. The population measured by a
spectrophotometer.
was grown in a commonly used nutrient growth
medium, but the pH of the medium was adjusted to
be pH 6.25.
Visualizing population density versus time
The bacteria population size was recorded every 16 minutes, giving us
precisely the snapshots we described in the discrete dynamical system
introduction. We'll let t denote time rescaled so that the time between
snapshots is one unit of time. Then, the above snapshots occurred at times
t=0,1,2,3,4,5. Time t=3, for example, indicates the actual time of 3×16=48
minutes.

The data consists of measurements of bacterial population density at each


time point. In our discrete dynamical system model, this population density
will be the single state variable in our state space, and we'll denote the
bacterial density by B. Then Bt is the snapshot of bacterial density at time
point t. With this notation, we can write the data from the above table as
B0=0.022, B1=0.036, B2=0.060, etc.
Visualizing population density versus time
With the state space and set of times determined,
the remaining step in developing the dynamical
system model is to determine the time evolution
rule. This step, of course, is the most difficult part,
as it must capture the dynamics of the system.
Sometimes, one has a sufficiently explicit
description of the dynamics so that one can
immediately write down a reasonable evolution
rule. In most cases, developing an equation that
captures key features of the underlying dynamics
will require multiple iterations and an effort to
look at both the experiment and the
mathematical system from different angles.
Visualizing population density versus time
Time Population Pop Change/U
We begin by looking at the plot of bacteria tt Density nit Time
BtBt Bt+1−BtBt+1-
density Bt versus the rescaled time t, above. It's Bt

immediately clear that the density is increasing 0 0.022 0.014


1 0.036 0.024
faster and faster as time progresses. Dynamics is
2 0.060 0.041
all about the change in the state variables, so let's 3 0.101 0.068
focus in on the change in Bt at each time step. 4 0.169 0.097
The population change per unit time is simply 5 0.266
Bt+1−Bt. These changes are, for example,
B1−B0=0.014 and B2−B1=0.024. Calculating all
these population changes, we add them to our
table of results. From now on, let's just use the
rescaled time t.
Visualizing population change versus time
We can visualize these dynamics by plotting
the population change as a function of time,
below. As we observed from the original
plot, the population change increases as a
function of time.
Visualizing population change versus time
If we stare at the above plot, we can observe that the population change
increases faster than linearly. In other words, the above points don't increase
as a straight line, but the trail of points tends to bend upward as time
increases. It doesn't look like we can make a good dynamical systems model
by finding some slope m and a constant c and writing Bt+1−Bt=mt+c. Lines are
one of the simplest types of equations, so we often want to find a nice simple
linear relationship among variables. But, here, it looks like it won't be able to
do that.
Visualizing population change versus time
If all we had were these points and we forgot the experimental system that
they came from, we might start employing some more mathematical tricks to
find a nice equation for these points. We might think that, since it doesn't
look like the points fall on a line, maybe they fall on a parabola. In fact, we
could find an equation for a parabola that gets fairly close to these points.

But, before we start going down that road, we need to stop and remember
the system we are trying to model. Our purpose in modeling the data is not
simply to find an equation that matches the data points. Our goal is a model
that captures the mechanism of the physical system we modeling. To achieve
a model that captures the key features of bacteria population growth, we
need to go back and recall what we know about how bacteria reproduce.
Visualizing population change versus time
As mentioned at the outset, bacteria cells reproduce by dividing into two
cells, and each cell takes about the same time to mature and divide. If our
snapshots occurred at the same interval as this dividing time, the population
size would double at each time step.

Looking at the above data, it doesn't appear that the population is doubling at
each time step. If it doubled, we would have expected that the change
Bt+1−Bt would be equal to the previous population size Bt. In all cases, the
change Bt+1−Bt is substantially less than Bt.
Visualizing population change versus time
What went wrong here? Let's imagine the bacteria cells are indeed dividing in
half after a certain period of time. The above snapshots were taken every
sixteen minutes. If each bacterium took sixteen minutes to mature and divide,
then we'd expect the population to double each time step of sixteen minutes.
Since the population growth seems to be slower, it's reasonable to assume
each bacterium takes longer than sixteen minutes to mature and divide.
Fitting a linear model to population change versus density

If cells are taking longer than a time step to mature and divide, a reasonable
assumption is that a certain fraction of the cells divide each time step. The
population change, Bt+1−Bt, should reflect the number of cells that divided
between the snapshot at time t and the snapshot at time t+1. Therefore, if
cells are dividing at a constant rate, then we expect that the population
change, Bt+1−Bt, should be some fixed fraction of the population Bt. In other
words, we expect that the model should be of the form
Bt+1−Bt=rBt (1)
for some constant r.
Fitting a linear model to population change versus density

We can test this hypothesis by plotting the data


a little differently. Instead of plotting
population change versus time, as we did
above, we should plot the population change,
Bt+1−Bt, versus the population size Bt. If
equation (1) holds, then the resulting points
should lie on a line that goes through the
origin. The slope of the line will give us the
constant r, which is the fraction of the bacteria
that divided in the time step.
Fitting a linear model to population change versus density
As shown in the above plot, the first four data points of population change versus density are
pretty much aligned along a line through the origin. The last point doesn't seem to be exactly
in line with the first four, but is a little lower. For now, let's not worry about the last point and
focus on the initial growth represented by the first four points

Substituting r=2/3 into equation (1), we arrive at the following evolution rule for our initial
discrete dynamical system model describing bacterial population growth:
Bt+1−Bt=23Bt.(2)
In words equation (2) says that the growth during the tth time interval is 23 times Bt, the
bacteria present at the beginning of the period. The number 23 is called the relative growth
rate, i.e., the growth per time interval is two-thirds of the current population size. More
generally, one may say that a fixed fraction of cells divide every time period. (In this instance,
two-thirds of the cells divide every 16 minutes.)

You might also like