An Invitation to Biomathematics

Chapter 1

Exercise
1-2.
........................
Because time is measured in decades elapsed since 1800, the value of
t for the year 3000 is t 120. Substituting this value for t in Eq (17)
gives Pt 5:3e0:297120 1:594
1016 . This prediction cannot
be realistic, because if we divide the earths land surface (estimated to
be approximately 150, 000, 000 km2) by this number, it would allow
less than 0.01 squared meters per person.

Solutions

Exercise
1-4.
........................
In this case, socioeconomic, cultural, and behavioral factors could play
a role in addition to environmental factors and availability of
resources.

Exercise
1-6.
........................
a) When P(0) 0, P(t) 0 for all values of t.
b) When 0 < P(0) < K, P(t) is increasing toward its carrying capacity
K and K is a horizontal asymptote for the graph of P(t).
c) When P(0) K, P(t) K for all values of t.
d) When P(0) > K, P(t) is decreasing toward its carrying capacity K
and K is a horizontal asymptote for the graph of P(t).
P(t)

Exercise
1-8.
........................

p3

a) The equilibrium state 0 is unstable, while K is a stable

equilibrium.

p2

b) The equilibrium points are p1, p2, and p3: p1 and p3 are stable
equilibria, while p2 is unstable. Sample trajectories for the
given initial conditions are given in Figure 1-2.

p1

0
FIGURE 1-2.
Time trajectories for Exercise 1-8, part b.

c) The equilibrium point p1 is neither stable nor unstable. If P(0) >

p1, P(t) will be increasing, because dP/dt > 0, and thus is moving
further away from p1. If P(0) < p1, P(t) will also be increasing,
because dP/dt > 0, and thus approaching the level p1.

Exercise
1-10.
..........................
a) The equilibrium states are 0 and 400: 0 is unstable and 400 is
stable.

Chapter 1

An Invitation to Biomathematics

b) 82 is an unstable equilibrium and 318 is a stable equilibrium.

c) From the graph in Figure 1-18 in the text, observe that
i) When P(t0) 50,

dP
< 0 and the population will decrease;
dt

dP
> 0 and the population will increase;
dt
dP
< 0 and the population will decrease.
When P(t0) 400,
dt

ii) When P(t0) 150,

f (P) =

dP
dt

d) The original graph will be shifted further down. Compared with

the graph from part (a), it will be shifted downwards by the
amount of cattle sold per year.
h

0
400

FIGURE 1-5.
Determining the maximal number of cattle sold
per year in the context of Exercise 1-10, for which
a stable equilibrium exists.

e) To answer this question, we need to find the distance h as shown in

Figure 1-5, such that the graph only touches the horizontal axis. As
long as the number of cattle sold per year is strictly smaller than h,
the graph will be crossing the horizontal axis in two points, thus
guaranteeing the existence of stable equilibria. To find h, notice that
it is equal to the vertical coordinate of the maximal value of the
function f (P). To find this maximum, we have to solve f (P) 0;
d2 P
that is 2 0. Computing this derivative gives
dt
d2 P
0:004P  8 0. Thus, the maximal value occurs at P 200,
dt2
and h f 200 0:002400  200200 80. Any number of cattle
sold per year which is less than this number guarantees the
existence of a stable equilibrium. Therefore, if 79 or fewer cattle are
sold per year, a stable equilibrium will exist.
f) 80.

Exercise
1-12.
..........................
a) The equilibrium states are found by solving P
solution. When P 6 0, we need to solve 1

2
P. P 0 is one
1 PK

2
, which gives P K.
1 PK

2
P0
> 1. In the
P > P0 because 1
P0 0
K
1 K
same way, Pn1 > Pn for any n 0, 1, 2, . . . . Thus, the population
size will be increasing.

b) When 0 < P(0) < K, P1

c) Denote the limit P lim Pn , n !1. Taking limits on both sides of

2
2
2
Eq. (1-27) gives: P lim Pn lim
P
lim Pn
P.
Pn n
limPn
1 KP
1 K
1 K

Solutions

An Invitation to Biomathematics

Because 0 < P(0) and because the population size increases with time,
P lim Pn 6 0. This, combined with the above calculations, implies
2
that the limit P satisfies the equation 1
. Solving this equation
1 KP
yields P K.

Exercise
1-14.
..........................


Pt  D
may change over time, thus causing
K
dP
to change. As a result, oscillations in the
the sign of the derivative
dt
solution trajectories may develop.

The sign of the term

1

a) To see this, assume that the population size is increasing toward its
carrying capacity K over the interval (t1  D; t1 ), and that at the moment
t1 the population reaches its carrying capacity (that is, P(t1) K). Then,
Pt1  D
dP
> 0, causing
> 0: At t1, the
because P(t1 D) < K, 1 
K
dt
population will be increasing and the time trajectory will thus cross over
the carrying capacity K, as in Figure 1-22 of the text.
b) The population P(t) will continue to increase in the interval
Pt  D
t1 ; t1 D, because for any t1 < t < t1 D; 1 
> 0, and
K
dP
thus
> 0.
dt
Pt  D
Pt1
dP
1
0,
0 and thus the
K
K
dt
population P(t) stops growing at t t1 D. Also, because P(t1) K and
P(t) was shown to be increasing over the time interval t1 ; t1 D, this
shows that P(t1 D) > K.
c) When t t1 D, 1 

When t > t1 D, but t is still close to t1 D, 1 

Pt  D
< 0, causing
K

dP
< 0. For time instances slightly larger than t1 D, the population
dt
size will be decreasing. If t2 is the first time after t1 at which
P(t2) K, P(t) will be decreasing over the interval t1 D; t2 D:

Exercise
1-16.
..........................
From the graph in Figure 1-24, the half life can be determined as the
value of t at which the concentration C(t) 2. Estimated from the graph,
this value appears to be close to 3.5. Analytically, using the given
ln2
equation, we obtain t
3:4657:
0:2

An Invitation to Biomathematics

Chapter 2

Exercise
1-18.
..........................
Thirty minutes after the last dose was given, the concentration will be
Ce10:5r Ce8:5r Ce6:5r Ce4:5r Ce2:5r Ce0:5r :

Exercise
1-20.
..........................
From Eq. (1-42) in the text, with C 10 mg/ml, r 0.1 hours-1,
C
10
80:1
T 8 hours, R Tr
8:15966 mg/ml.
e 1 e
1

Exercise
1-22.
..........................
Choose the initial dose to be just under C0 MTC. Subsequent doses will
all be equal to C MTC MEC. The time between all doses should be
1 MTC
, where r is the elimination constant for
calculated as T ln
r MEC
the drug.

Exercise
1-24.
..........................
The following values would be appropriate:
STARTTIME 0
STOPTIME 7
DT 0.05

Chapter 2

Exercise
2-2.
........................
b
dI
> 0, the rate of change
is given by
a
dt
b
the logistic Eq. (2-3) with K N  . It is known from Chapter 1
a
b
then that when I0 < N  , the function I(t) is increasing toward
a
b
b
K N  . That is, lim It N  , and lim St
t!1
t!1
a
a


b
b
lim N  It N  lim It N  N 
. Because I(t)
t!1
t!1
a
a
S(t) N, the fact that I(t) increasing implies that S(t) is decreasing.

a) Recall that when N 

Solutions

An Invitation to Biomathematics

b
> 0, we know that the solution of the
a
logistic Eq. (2-3) has a time trajectory I(t) that decreases towards
b
b
K N  . Therefore, as in part a), lim It N  .
t!1
a
a
b
Because S(t) I(t) N, this is equivalent to N  lim It
t!1
a
b
lim N  It lim St, showing that lim St .
t!1
t!1
t!1
a



b dI
b
c) When I0 N  , jt0 aI0 N   I0 0 and the
a dt
a
b
level of infectives would remain at It N  for all values
a
of t > 0.

b) When I0 > N 

Exercise
2-4.
........................
a) Heuristically, the size of the susceptible group should be
decreasing, because the model assumes complete immunity after
recovery. Thus, once an infective leaves the susceptible group,
there is no way for this infective to return to this group, and the
number of susceptibles S(t) should be decreasing. Mathematically,
dS
the same can be verified by examining the sign of the derivative .
dt
dS
Because
aSI < 0, the functions S(t) is decreasing.
dt
b) The size of the recovered group is initially zero and should be
increasing with the number of individuals who have recovered
from the infection. Once in R, no individual can leave the group,
because according to the model assumptions, recovery confers
permanent immunity. Mathematically, because the derivative
dR
bI > 0, the function R(t) is increasing.
dt
c) Because S(t) I(t) R(t) N for all values of t,
Thus aSI

dS dI dR

0.
dt dt dt

dI
dI
bI 0, and aSI  bI.
dt
dt

Exercise
2-6.
........................
dS dS=dt
a
dS
a

 S, we obtain
 dR. Integrating both
dR dR=dt
b
S
b
Z
Z
dS
a
sides yields

dR
S
b

Because

Chapter 2

An Invitation to Biomathematics

a
lnS  R C , where C is a constant of integration. Thus,
b
a
b

 RtC

St e

C b R0

S0 e e

eC e

a
b

 Rt

. For t 0, because R(0) 0, we obtain

a
b

 Rt

eC , which proves that St S0e

Exercise
2-8.
........................
dR
> 0 for all values of t, mathematically, this means that the
dt
function R(t) is increasing for all values of t. There are two possibilities.
Either lim Rt 1 or lim Rt L < 1. However, in the latter case,
When

t!1

t!1

L would be a horizontal asymptote for R(t), which would imply

dR
dR
lim
0. Because we know that lim
> 0, this leaves us with the
t!1 dt
t!1 dt
first possibility, namely lim Rt 1.
t!1

Exercise
2-10.
..........................
Possible scenarios may include:
1. The owl population goes extinct and the vole population grows
undisturbed in the absence of predators;
2. The vole population goes extinct because of a lack of resources, and
the owl populations follows because they have no prey to feed on;
3. The vole and owl populations reach equilibrium at levels that allow
for the two populations to coexist;
4. The vole and owl populations coexist but neither population is in
equilibrium. The balance is maintained by sustained oscillations in
the population sizes of predator and prey similar to those depicted
in Figures 2-15, 2-16, and 2-17.

Exercise
2-12.
..........................
a) The equilibrium states are the solutions of the system of equations
dV
f V; O aV  bV 2  gOV Va  bV  gO 0
dt
dO
gV; O dO eOV Od eV 0:
dt
Notice that when O 0, the equation

dO
0 is satisfied,
dt

regardless of the value of V. In the same way, when V

d
, the
e

Solutions

An Invitation to Biomathematics

dO
0 is satisfied, regardless of the value of O.
dt
dV
Substituting O 0 in the equation
0 gives V(abV) 0,
dt
a
a
which implies V 0 and V . Thus (0,0) and ( , 0) are
b
b
d
dV
equilibrium states. Substituting V in the equation
0 gives
e
dt



d
d
a  b  gO 0. Solving this equation for O implies
e
e



a bd
d a bd
, and establishes that
; 
O 
is also an equilibrium
g ge
e g ge
state.
equation

b) Computing the partial derivative for the functions f and g above

gives:
@f
a  2bV  gO
@V

0
1
0
1
@f @d a bdA
d
a
bd
bd
; 
a  2b  g@  A
@V e g ge
e
g ge
e
0

@f
gV
@O

@f @d a bdA
d
gd
; 
g 
@O e g ge
e
e

@g
eO
@V

0
1
0
1
@g @d a bdA
a
bd
ae  bd
; 
e@  A
@V e g ge
g ge
g
0

@g
d eV
@O

@g @d a bdA
d
; 
d e 0
@O e g ge
e

Exercise
2-14.
..........................
a) We begin by writing the general form of the equation describing
dN
rN; 0N. Because the
the change in the population N:
dt
resources are limited, the per capita rate of growth r(N,0) will have
to be decreasing as N is increases. Following arguments similar to
those in Chapter 1, the following logistic equation may be used to



dN
N
model the situation:
a 1
N, where K is the carrying
dt
K
capacity for the population N and a is its inherent per capita rate of
growth.
b) As in part a), if N 0, we can use the logistic equation



dP
P
c 1
P to describe the change in the size of population P.
dt
M

An Invitation to Biomathematics

Chapter 2

Here M is the carrying capacity for this population and c is its

inherent per capita rate of growth.
c) The equations from parts a) and b) are logistic equations describing
the growth of a single population, as introduced in Chapter 1.
d) For part a), K is the carrying capacity of population N, and a is its
inherent per capita rate of growth. For part b), M is the carrying
capacity of population P, and c is its inherent per capita rate of growth.

Exercise
2-16.
..........................
To determine the equilibrium states of the model described by Eqs. (2-15)
in the text, we need to determine the values of N and P for which
0
1
dN
N

bP
AN 0
a @1 
dt
K
0
1
dP
P gN A
c@1 
P 0:
dt
M
This leads to the following 4 possibilities:
1. N 0 and P 0;
2. N 0 and 1 

P gN
P
0 (that is, 1  0), giving the
M
M

equilibrium state N 0 and P M;

3. P 0 and 1 

N bP
N
0 (that is 1  0), giving the equilibrium
K
K

state N K and P 0;
N bP
P gN
0 and 1 
0, which yields the system of
K
M
equations

4. 1 

K  N  bP 0
M  P  gN 0:
Solving for N and P gives the fourth equilibrium state N
and P

K  bM
1  bg

M  gK
.
1  bg

Exercise
2-18.
..........................
In the model from Eq. (2-15), let g 0.51, K 2000, and M 1000. Prove
that for these values of the parameters, the equilibrium state (K, 0) is
asymptotically stable, regardless of the values of a > 0, b > 0, and c > 0.

Solutions

An Invitation to Biomathematics

As in Exercise 2-17, denote

0
1
dN
N

bP
AN f N; P
rN; PN a@1 
dt
K
0

dP
P gN A
kN; PP c@1 
P gN; P;
dt
M
and compute
@f
2a
ab
a N P
@N
K
K

@f
K; 0 a
@N

@f
ab
 N
@P
K

@f
K; 0 ab
@P

@g
cg
 P
@N
M

@g
K; 0 0
@N

@g
2c
cg
@g
c P N
K; 0 c  2cg:
@P
M
M
@P



a
ab
Then, J
, and
det J ac  2cg ac2g  1,
0 c  2cg
traceJ a c  2cg a  c2g  1. Because for g > 1/2,
det( J) > 0 and trace ( J) < 0 (independent of the values of a, b, and c),
the equilibrium state is asymptotically stable for g 0.51.

Chapter 3

Exercise
3-2.
........................
In this case, there are 3 sites that we can assume are independent
because they are on different chromosomes. The possibilities for the A
site are AA, Aa, aA, and aa; the possibilities for the B site are BB, Bb,
bB, and bb; and the possibilities for the C site are CC, Cc, cC, and cc.
a) (3/4)(3/4)(3/4) 27/64
b) (3/4)(3/4)() 9/64
c) (3/4) ()() 3/64
d) ()()() 1/64
e) In addition to the phenotypes from parts a)-d) the following are
possible: the A and C sites are dominant and the B site is recessive
(with probability 9/64); the B and C sites are dominant and the

10

An Invitation to Biomathematics

Chapter 4

A site is recessive (with probability 9/64); the B site is dominant

and the A and C sites are recessive (with probability 3/64); the
C site is dominant and the A and B sites are recessive (with
probability 3/64).

Exercise
3-4.
........................
Because females have two X chromosomes, the possible female
genotypes are: CC, Cc and cc with frequencies p2, 2pq, and q2,
respectively. Because males have one X and one Y chromosome, and
because the gene C is found on the human X chromosome only, the
possible male genotypes will be C and c, with frequencies p and q.

Exercise
3-6.
........................
With m 6 genes, there are N 2m 12 alleles and 212 4096 allele
combinations. Let k denote the number of contributing alleles.




1
1
0:000244
a)

12
4096
2



1
12! 1
12

0:193359
b)

12
5
5!7! 4096
2

12
12




1
12! 1

0:225586

12
6!6! 4096
2



1
12!
1
12

0:016113
d)

2
2!10! 4096
212
c)

12
6

Chapter 4

Exercise
4-2.
........................
HYPOTHESES
(1) The right hand of a person has a larger span;
(2) The spans of left and right hands are correlated across subjects;
(3) The dominant hand of a person has a larger span;
(4) The spans of dominant and nondominant hands are correlated
across subjects.

Solutions

An Invitation to Biomathematics

11

Study design and variables: This study has a repeated measures designeach
participant contributes two measurements of her left and right
hands. These measurements are also paired differentlydominant
versus nondominant hand. Thus, paired tests will be needed to address
hypotheses 1 and 3. These tests will use one-tail significance level because
hypotheses 1 and 3 are directional. Hypotheses 2 and 4 will be addressed
by directly correlating the measurements of the two hands of all
participants:
Hypothesis 1: Table 4-2 presents the average span of the left versus right
hand of the participants. A paired t tests asserts that the observed
distance is significant at p < .05.
Hypothesis 2: The measurement of left and right hands of the participants
are highly correlated (r 0.87; p < .001). This confirms that when a
person has one hand which is larger than average, it should be expected
that this difference will hold for the other hand.
Hypothesis 3: Table 4-2 presents the average span of the dominant versus
nondominant hand of the participants. A paired t tests asserts that the
observed distance is highly significant.
Hypothesis 4: The measurement of dominant and nondominant hands of
the participants are highly correlated (r 0.91; p < .001). This correlation
is apparently (although not statistically significantly) stronger than the
correlation between left and right hands of the participants.

FUTURE RESEARCH
It appears that the conclusions derived from the tests of dominant versus
nondominant hand are similar to, but stronger than, the conclusions
derived for right versus left hand, both in terms of differences in span
and in terms of correlations. Because for most participants in this study
the dominant hand was also the right hand, we can speculate that the
observed differences in the hand span are caused by the dominant hand
being larger. This speculation is reinforced by a close examination of the
data of the two participants who reported a dominant left handfor
both, their left hands had larger spans than their right hands. Thus,

Test of Hypothesis 1

Mean

Test of Hypothesis 3

Right Hand Span

Left Hand Span

Dominant Hand Span

Non-Dominant Hand Span

19.10 cm

18.73 cm

19.20 cm

18.63 cm

Difference (SE)

0.37 (0.19)

0.57 (0.15)

Paired t test; P

t 1.94; 1-tail p .036

t 3.77; 1-tail p .001

TABLE 4-2.
Results for Exercise 4-2.

An Invitation to Biomathematics

12

Chapter 6

further research is needed to determine that the difference in hand span

is entirely accounted for by dominant versus nondominant hand, and
the distinction between right and left hand is not important. Such
research would require recruiting a case-controlled sample of an equal
number of participants who are right-handed and left-handed. If such a
sample is well-balanced, we should expect no difference between the
span of left versus right hands, and no correlation between left and right
hands across the participants. However, we should expect significant
differences and correlations between dominant and nondominant hands.

Chapter 5

Exercise
5-2.
........................
First, we need to compute f(1.8), f(3.9), f(6.25), and f(24.6), from the
function f BG 1:774 lnBG1:0329  1:8707. Next, for each of these
values, we need to compute the risk function as rBG 10 f BG2 .
For example, f 3:9 1:774 ln3:91:0329  1:8707 0:87
and r3:9 10 f 3:92 7:58. The complete set of values is presented
in Table 5-1.

Exercise
5-4.
........................
BG

f(BG)

r(BG)

1.80

2.29

52.62

3.90

0.88

7.74

6.25

0.00

0.00

24.60

2.58

66.81

TABLE 5-1.
Results for Exercise 5-2.

The results in Table 5-1 of the text illustrate that LBGI and HBGI appear
to be capable of capturing the differences between the two data sets in
Figure 5-10. However, because the data comes from only 2 subjects
(1 with T1DM and 1 with T2DM), the results cannot be used as
a validation of the model, as it is likely that some (or all) of the
observed phenomena can be attributed to chance. A credible validation
should include a large number of T1DM and T2DM subjects to minimize
the element of chance caused by unavoidable differences in the patients
BG control.

Chapter 6

Exercise
6-2.
........................
Consider, for example, the following two series of 10 RR intervals each,
given in milliseconds:

Solutions

An Invitation to Biomathematics

A f658:3; 666:7; 659:3; 666:7; 697:7; 652:2; 666:7; 681:8; 705:9; 666:7g
B f659:0; 674:2; 659:3; 666:7; 625:0; 652:2; 666:7; 674:2; 631:6; 674:2g
Series A has median value of 666.7 and series B has median equal to 663.
The standard deviations of the RR intervals in series A and B are
identical, both equal to 17.51 milliseconds, as presented in the table
below. However, series A (with transient decelerations) has a much
higher R2 that R1, while series B (with transient accelerations) has much
higher R1 than R2. These results are summarized in Table 6-1.
Standard Deviation

R1

R2

Series A

17.51

33.36

272.56

Series B

17.51

257.63

40.37

TABLE 6-1.
Results for Exercise 6-2.

Exercise
6-4.
........................
Similar to the LBGI and HBGI defined in Chapter 5 that provide
cumulative measures for the risk for hypoglycemia and hyperglycemia,
the measures R1 and R2 defined in Eq. (6-2) of the text provide
a cumulative measure for the risk associated with accelerations
and decelerations for a data set of RR measurements. In fact, the
mathematical definition of R1 and R2 in Eq. (6-2) based on risk functions
is exactly the same as that for the LBGI and HBGI. The difference is in
the specific definitions of the risk functions rd(x) and ra(x), as outlined
in Exercise 6-3 above.

Exercise
6-6.
........................
Note that the sum in the definition of R1 is over the entire sequence
(n data points) and R1 is the average risk for RR accelerations for the
sequence. In the same way, R2 is the average risk for RR decelerations.
Further, recall that a data point that corresponds to a measurement
larger than the median, has ra(x) 0 and rd(x) > 0. Conversely, a data
point that corresponds to an RR measurement smaller than the median,
has ra(x) > 0 and rd(x) 0. Thus, for two sequences of equal length,
R1 is larger for the sequence with more accelerations and R2 is larger
for the sequence with more decelerations. Subsequently, the ratio
SA R2 =R1 will grow when there are more decelerations in the RR
sequence and will decrease when there are more accelerations.
This establishes (1) and (2). To verify (3), notice that because both ra(x)
 0 and rd(x)  0, both R1 and R2 are non-negative, and thus
SA R2 =R1  0. Finally, when the distribution for an RR sequence is
perfectly symmetric, for any data point x with a non-zero value for ra(x)
there will be another data point y in the sequence with rd(y) ra(x).

13

14

Chapter 7

An Invitation to Biomathematics

This will cause the values of R1 and R2 to be exactly the same, leading to
SA R2 =R1 1. This establishes the condition (4).

Exercise
6-8.
........................
The standard deviation for S1 is SD 0.527. Thus t r  SD
0:20:527 0:1054. Because the standard deviations for the sequences
S2 and S1 are the same (since they both have five 0s and five 1s),
the tolerance value for S2 is also t 0.1054.

Exercise
6-10.
..........................
Calculate R1, R2 and the Sample Asymmetry (SA). The SA of RRI series
with increased proportions of decelerations will be higher than 1, while
the SA of RR series with increased proportions of accelerations will be
lower than 1. For example, in the data of Exercise 6-2, for series A the
SA8.02, while for series B the SA0.16.

Chapter 7

Exercise
7-2.
........................
Applying the law of mass action to the equation Hbn nO2 $ Hbn O2 n ,
we obtain Hbn O2 n  kHbn O2 n , where k is the association constant.
Then for the fractional saturation we calculate:
Y

Hbn O2 n 
kHbn O2 n
kO2 n

;
n
Hbn  Hbn O2 n 
Hbn  kHbn O2 
1 kO2 n

as in Eq. (7-9) of the text.

Exercise
7-4.
........................
a) Differentiating Eq. (7-10) in the text with respect to the oxygen
concentration [O2] gives
dY
n k O2 n1 1 k O2 n  k O2 n n k O2 n1

dO2 
1 kO2 n 2

n k O2 n1 k O2 n n k O2 n1  k O2 n nkO2 n1

1 k O2 n 2

n k O2 n1
1 k O2 n 2

b) When the concentration of [O2] decreases to zero, we obtain

from part a) that the denominator 1 kO2 n 2 ! 1 when O2  ! 0.

Solutions

An Invitation to Biomathematics

For the numerator, because n > 1 and, thus, n1 > 0, we obtain
dY
lim n k O2 n1 0.
n k O2 n1 ! 0, and thus lim
O2 !0 dO2 
O2 !0

Exercise
7-6.
........................
In the context of the model used to obtain Eq. (7-6) in the text, we need to
examine myoglobinoxygen binding. Because myoglobin binds one
oxygen molecule, only two binding species will be present in the
solution: unbound myoglobin and myoglobinoxygen complex. Recall
that the binding polynomial X is defined as the sum of the
concentrations of all the binding species present in the solution. In this
case, therefore, we obtain X M MO2  M Ka MO2  where M
denotes myoglobin and Ka is the myoglobin-oxygen association constant.
Expressing the units of myoglobin concentration as a fraction of the
unoxygenated myoglobin concentration (as was done to obtain the
binding polynomials in Eqs. (7-27)), we obtain X 1 Ka O2 .
Further, because there is a single binding site, Y N, and from
Eq. (7-20) we obtain Y N O2 
because

@ lnX O2  @X


.
@O2 
X @O2 

@X
Ka O2 
Ka , we obtain Y
.
@O2 
1 Ka O2 

Chapter 8

Exercise
8-2.
........................
For the model Y Ga; X aX, the sum of squared residuals measure is
n
X
X
X
SSRa
r2i
Yi  Ga; Xi 2
Yi  aXi 2 :
i1

The value for a that minimizes this expression can be found as the
solution of the equation:
X
X
X
@SSR
2
Xi Yi  aXi  0; or
Xi Yi  a
Xi2 0:
@a
i
i
i
Therefore, for the linear model Y Ga; X aX, the least squares
estimate for the parameter a is given by the formula
X

Xi Yi
a X 2 :
Xi
i

15

16

An Invitation to Biomathematics

Chapter 8

Exercise
8-4.
........................
@SSRr; c
@SSRr; c
and
for the
Computing the partial derivatives
@r
@c
X
function SSRr; c
Yi  Gr; c; Xi 2 gives
i1

X
@SSRr; c
@Gr; c; Xi
2
Yi  Gr; c; Xi 
@r
@r
i

X
@SSRr; c
@Gr; c; Xi
2
, showing that
Yi  Gr; c; Xi 
@c
@c
i
2
3
@SSRr; c
6
7
@r
6
7
T 
6
7.
P Y 26
7
4 @SSRr; c 5
@c
Next, because e can be expressed as e PT P1 PT Y (see Example 8-5
in the text), PT P1 PT Y 0 when e 0. This means that either
PT P1 0 or PT Y 0. However, PT P1 cannot be the zero matrix
because it is invertible. Thus, PT Y 0, implying that
3
2
@SSRr; c
7
6
@r
7
6
@SSRr; c
@SSRr; c
7
0 and
0
PT Y 26
6 @SSRr; c 7 0. Thus,
@r
@c
5
4
@c
when e 0. This shows that the GaussNewton procedure described in
the text produces the least squares values for the parameters r and c.

Exercise
8-6.
........................
(a) For Y Ga; b; c; d; e; f ; X aX5 bX4 cX3 dX2 eX f , the
first-order derivatives with respect to the parameters are
@Ga; b; c; d; e; f ; Xi
@Ga; b; c; d; e; f ; Xi
Xi5 ,
Xi4 ,
@a
@b
@Ga; b; c; d; e; f ; Xi
@Ga; b; c; d; e; f ; Xi
Xi3 ,
Xi2 ,
@c
@d
@Ga; b; c; d; e; f ; Xi
@Ga; b; c; d; e; f ; Xi
Xi , and
1. Therefore, all
@e
@f
higher order derivatives for the model will be equal to zero,
leaving only the first-order terms in Eq. (8-38) in the text,
which corresponds to a linear model.
(b) Analogously to part (a), any model Y Gam ; am1 ; . . . a1 ; a0 ; X
am Xm am1 Xm1 . . . a1 X a0 where the degree of the

Solutions

An Invitation to Biomathematics

polynomial Y am Xm am1 Xm1 . . . a1 X a0 is a fixed

positive integer, is a linear model. This holds because
@Gam ; am1 ; . . . ; a1 ; a0 ; Xi
Xik , for all k 0, 1, . . ., m, and all
@ak
higher order derivatives with respect to the parameters are zero.

Exercise
8-8.
........................
(a) This follows from direct matrix multiplication that establishes
Pe Y
(b) Multiplying both sides of the equation from part a) by the
transposed matrix PT gives PT Pe PT Y . Multiplying both
sides by the inverse PT P1 leaves to PT P1 PT Pe
PT P1 PT Y or, in simplified form e PT P1 PT Y .

Chapter 9

Exercise
9-2.
........................







2pt 5
2pt
2pt
2p sin
Notice that ht 5 sin
sin
5
5
5
ht, which shows that h(t) is periodic with period T 5.






2pt 5
2pt
In the same way, rt 5 cos
2p
cos
5
5


2pt
cos
rt, which shows that r(t) is periodic with period T 5.
5

Chapter 10

Exercise
10-2.
..........................
We follow the steps outlined in the hint to compute
Ct h  Ct
:
C0 t lim
h!0
h
(1)

Zth
Zth
athz
Ct h Sze
dz Szeah eatz dz
1

1

ah

Zth
Szeatz dz:
1

17

18

Chapter 10

An Invitation to Biomathematics

(2)

Ct h  Ct e

ah

Zth
Zt
atz
Sze
dz  Szeatz dz
1

1

Zt
Zth
Zt
atz
ah
atz
Sze
dz e
Sze
dz  Szeatz dz

eah

1

1

Zth
Zt
atz
ah
ah
e
 1 Sze
dz e
Szeatz dz:
1

(3) Ct h  Ct
eah  1

h
h

Zt
Zth
eah
atz
Sze
dz
Szeatz dz
h

1

eah  1
eah
Ct
h
h

Zth
Szeatz dz:
t

1
(4) Using that lim
h!0 h

Zth
eah  1
a;
Szeatz dz St and lim
h!0
h
t

we obtain
Ct h  Ct
eah  1
eah
lim
lim
Ct lim
h!0
h!0
h!0 h
h
h

Zth
Szeatz dz
t

aCt St:

Exercise
10-4.
..........................
In the context of Eq. (10-3); that is, Ct C0  S=aeat S=a; the
problem specifics are as follows:
Half-life t0 12 minutes: Because
t0

ln 2
ln 2 ln 2
;a
0:0578min1 :

a
t0
12

S(t) S const.
C(0) 0
C(3) 500 ng/ml
We seek the value of S for which
500 C3 C0  S=aeat S=a; which after substitution of the above
values gives

Solutions

500 

An Invitation to Biomathematics

S
S
S
e3
0:0578

1  e3
0:0578 ; and
0:0578
0:0578 0:0578
S

5000:0578
181:5 ng=ml=min:
1  e3
0:0578

Exercise
10-6.
..........................
In this case, Eq. (10-1) becomes
dCA
aCA t SA t aCA t SA;basal aFup;down CB :
dt
Because the maximal and minimal concentrations of A are achieved
dCA
0; this means that the minimal and maximal concentrations
when
dt
are achieved when aCA t SA;basal aFup;down CB 0 or,
SA; basal aFup;down CB
:
equivalently, CA t
a
(1) As the maximal value of Fup(doun) is 1, CA t 
values of t, and therefore CA;max
a aCA;max  SA;basal :

SA;basal a
: From here,
a

(2) As the minimal value of Fup(doun) is 0, CA t 

of t, and therefore CA;min

SA;basal a
for all
a

SA;basal
for all values
a

SA;basal
. From here, SA;basal aCA;min:
a

Exercise
10-8.
..........................
(a) For simplicity, denote CA(t) x(t) x, CB(t) y(t) y, and
Fyt  DB a
Gx b

1
1
a
CB t  DB =TB nB 1
yt  DB =TB nB 1

CA t=TA nA
x=TA nA

b
:
CA t=TA nA 1
x=TA nA 1

Notice that F(y) is a down-regulatory Hill function and is,

therefore, monotone decreasing as a function of y. The function G(x) is
and up-regulatory Hill function and is, therefore, monotone increasing
as a function of x. Also, both functions are strictly positive; that is
F(y) > 0 for all values of y and G(x) > 0 for all values of x.
(1) With this notation the system from Eqs. (10-14) becomes
dx
x0 ax Fyt  DB
dt
dy
y0 by Gx
dt

19

20

Chapter 10

An Invitation to Biomathematics

(2) The equilibrium states of this system of equations are the pairs
of time-independent constants x and y satisfying the system:
ax Fy 0
by Gx 0
or, equivalently, the system
ax Fy
by Gx:
(3) Expressing y Gx=b from the second equation, and
substituting into the first yields x FGx=b=a: The solutions
of this equation then will correspond to the intersection points
of the graphs of the functions z FGx=b=a and z x: We
next show that there is a unique positive intersection point.
Because the function G(x) is monotone increasing, and the
function F(y) is monotone decreasing, the composition
FGx=b=a is also because a > 0 and b > 0 monotone
decreasing as a function of x. Also, because at x 0,
FG0=b=a > 0; the graph of the function z FGx=b=a
is above the x axis for x 0 and decreases with x. As the
graph of z x begins at z 0 for x 0 and increases with
x, the two functions will cross at a single point. This proves
that the equation x FGx=b=a has only one positive
solution.
Then, because y is calculated as y G(x)/b, we obtain that the
system of differential equations (14) has an equilibrium state
that is unique.
(b) When DB 0; the function Fy

1
1

.
CB t=TB nB 1 y=TB nB 1

(4) The Jacobian of the system of differential equations from

part (1) above is


J

a
G0 x


F0 y
;
b

trJ a b < 0; and

det J ab  F0 yG0 x:
Because Fy is strictly monotone decreasing, F0 y < 0 for all
values of y. Because the functions G(x) is strictly monotone
increasing, G0 x > 0: Then F0 yG0 x < 0 and, thus,
detJ ab  F0 yG0 x > 0: This proves that when DB 0; the
single equilibrium state for the system of Eqs. (10-14) is
asymptotically stable.

Solutions

An Invitation to Biomathematics

Exercise
10-10.
.............................
(a) Passing to a limit for TA ! 0 in the second equation of Eqs. (10-17)
b
gives 0 <  e  CB t  b=b e for sufficiently large t. Because
b
these inequalities hold for any sufficiently small e > 0; this
proves that when TA decreases, the pulsatility of hormone B
vanishes.
(b) Passing to a limit for TB ! 1 in the first equation of Eqs. (10-17)



b nB
a
a
! 0) 0 <  e  CA t  e for
gives (because
bTB
a
a
sufficiently large t. Because these inequalities hold for any
sufficiently small e > 0; this proves that when TB increases, the
pulsatility of hormone A vanishes.
To provide a heuristic explanation for the observed changes in model
behavior, we note that if a system change results in a separation of the
concentration of one of the hormones from its action threshold, the
oscillations will vanish, because the system will be less sensitive to
variations in the concentrations of this hormone. In our particular case,
the hormone concentrations in the circulation are limited. Therefore, the
concentration of A and B will be disassociated from their corresponding
action thresholds if either TA ! 1 or TB ! 1 and the oscillations will
disappear. In addition, under no circumstances can B fully block the
release of A. Therefore, when TA decreases TA ! 0; we will see again
a separation of the concentration of A from its action threshold. On the
other hand, if TB ! 0 the concentration of B may be able to drop below
this threshold, because the potential of B to block the secretion of A will
increase, which would lead to reduced secretion of A and, hence, a lower
secretion of B. Therefore, the restorative process that generates the
oscillations in the model would still operate.

Chapter 11

Exercise
11-2.
..........................
COSIN2NL: The results for the ARFILTERed time series are close
to those for the original time series. In both cases, the period is
estimated to be approximately 23.3  0.19 hours. Thus, filtering
the time series before the analysis does not appear to be necessary
or beneficial.
FFT-NNLS: For this analysis, too, the results for the ARFILTERed
time series are close to those for the original time series. In both cases,

21

Chapter 12

An Invitation to Biomathematics

24
20
Time of Acrophase

22

16
12
8
4
0

Day

FIGURE 11-1.
A plot depicting the daily time of acrophase for a time series with a period of 20 hours.

four periodic components were identified, with the circadian period

estimated at 23.4  0.31 for the original time series and 23.4  0.28 for
the AIRFILTERed series. Thus, filtering the time series before the
analysis does not appear to be necessary or beneficial.
PHASEREF: In this case, the results for the ARFILTERed time series and
for the original time series differ significantly. The mean period values
for the ARFILTERed time series ( SD, with SEM values in parentheses)
presented next to the graphs in Figure 11-29 are a lot closer to the actual
simulated values of 24 hours than those next to the graph for the noisy
time series in Figure 11-25. This is a consequence of noise confounds
creeping in for the original time series, beginning at about 100 hours of
x-axis time and manifesting consistently beyond about 200 hours of
x-axis time (Figure 11-26). In contrast, for the smoothed time series, there
are fewer instances in which values for period estimates are considerably
shorter than 24 hours and they do not appear until about 217 hours of
x-axis time (Figure 11-30). As a result, the period estimates are more
accurate for the ARFILTERed time series, making the use of ARFILTER
before PHASEREF beneficial.

Chapter 12

Exercise
12-2.
..........................
Let xi xi1 ; xi2 ; . . . ; xin and xk xk1 ; xk2 ; . . . ; xkn be the i-the and k-th
row of the gene expression matrix. We want to show that the
dissimilarity measure
v
uX
u n
dxi ; xk t
xij  xkj 2
j1

satisfies conditions (1) through (3).

Solutions

An Invitation to Biomathematics

(1) Because dxi ; xk is chosen to be the positive value of the square

root, the condition dxi ; xk  0 is satisfied.
(2) dxi ; xk dxk ; xi is obvious, because
v v
uX
uX
u n
u n
2
t
xij  xkj t
xkj  xij 2 :
j1

j1

3) Let xs xs1 ; xs2 ; . . .; xsn be any other vector. Then we want to show
that
v v
uX
uX
u n
u n
dxi ; xs dxs ; xk t xij  xsj 2 t xsj  xkj 2
j1

j1

v
uX
u n
 t xij  xkj 2 dxi ; xk :
j1

We will use the following fact, known as the CauchySchwartz inequality:

If a1 ; a2 ; . . . ; an and b1 ; b2 ; . . . ; bn are two sets of real numbers, then
pq
a1 b1 a2 b2    an bn  a1 2 a2 2    an 2 b1 2 b2 2    bn 2 , or,
r

Xn
Xn rX
n
equivalently,
a
b

a
b.
j1 j j
j1 j
j1 j
We now use this inequality to show that dxi ; xk  dxi ; xs dxs ; xk :
d2 xi ; xk

n
X

xij  xkj 2

j1

n
X

xij  xsj xsj  xkj 2

j1

n
X
xij  xsj 2 xsj  xkj 2 2xij  xsj xsj  xkj 
j1

n
X

xij  xsj 2

j1

n
n
X
X
xsj  xkj 2 2 xij  xsj xsj  xkj
j1

j1

 applying CauchySchwartz
vv
uX
uX
n
n
X
X
u n
u n
2
2
xij  xsj
xsj  xkj 2t
xij  xsj 2 t
xsj  xkj 2
j1

j1

j1

j1

d2 xi ; xs d2 xs ; xk 2dxi ; xs dxs ; xk dxi ; xs dxs ; xk 2

Thus, d2 xi ; xk  dxi ; xs dxs ; xk 2 , which yields
dxi ; xk  dxi ; xs dxs ; xk .

23

24

An Invitation to Biomathematics

Chapter 12

Exercise
12-4.
..........................
Let xi xi1 ; xi2 ; . . . ; xin and xk xk1 ; xk2 ; . . . ; xkn be the i-the and k-th
row of the gene expression matrix. We want to show that the
dissimilarity measure
dxi ; xk maxj jxij  xkj j
satisfies conditions (1) through (3).
(1) Because the absolute value is always non-negative and the
dissimilarity measure is defined as the maximum of the absolute
values, dxi ; xk  0 is satisfied.
(2) dxi ; xk dxk ; xi is obvious, because for any two numbers a and
b, ja  bj jb  aj, and thus
dxi ; xk maxj jxij  xkj j maxj jxkj  xij j dxk ; xi :
3) To prove that dxi ; xk  dxi ; xs dxs ; xk , we will use the
well-known triangle inequality for real numbers which states
that for any real numbers a, and b, jaj jbj  ja bj.
Now,
dxi ; xs dxs ; xk maxj jxij  xsj j maxj jxsj  xkj j 
maxj jxij  xsj j jxsj  xkj j  maxj jxij  xkj j dxi ; xk ;
establishing that dxi ; xk  dxi ; xs dxs ; xk . The last inequality in the
chain of inequalities above follows from the triangle inequality, because
jxij  xsj j jxsj  xkj j  jxij  xsj xsj  xkj j  jxij  xkj j: