Physical Biology of the Cell

Hints to the Problems

Rob Phillips, Jane Kondev, Julie A. Theriot and Hernan G. Garcia

January 18, 2013

Contents

Preface v

2 Biological Structures: Rulers at Many Different Scales 7

4 Who: Bless the Little Beasties 9

5 Mechanical and Chemical Equilibrium in the Living Cell 11

6 Entropy Rules! 13

7 Two-State Systems: From Ion Channels to Cooperative Binding 15

8 Random Walks and the Structure of the Macromolecules of the

Cell 17

9 Electrostatics for Salty Solutions 21

10 Beam Theory: Architecture for Cells and Skeletons 23

11 Biological Membranes: From the Golgi Apparatus to Membrane

Fusion 25

13 Random Walks: A Statistical View of Biological Dynamics 29

15 Rate Equations and Dynamics in the Cell: Cytoskeleton Under

Construction 33

16 Dynamics of Molecular Motors 35

17 Biological Electricity and the Hodgkin-Huxley Model 37

iii
iv CONTENTS

18 Light and Life 39

19 Organization of Biological Networks 41

21 Sequences, Specificity, and Evolution 43

Preface

This document contains a series of hints to guide the reader with some of the
more difficult problems in the book. The goal of the hints is to give the reader
a better sense of what we are asking for in the problem since often the problem
statements can be complemented by extra information. Second, the goal of the
hints is to sketch a logical attack on the problem of interest.
If our readers find a specific problem obscure or deserving of a hint, please
let us know and we will consider adding a hint for the problem of interest.

v
Chapter 2

Biological Structures:
Rulers at Many Different
Scales

2.4. A feeling for the numbers: molecular volumes and masses.

(a) If you want to use actual PDB files of amino acids to calculate their volumes
you’ll have to look around on the Internet. A simple Google search for “amino
acid pdb” should give you plenty of relevant pages. One useful one, for example,
is http://wbiomed.curtin.edu.au/teach/biochem/tutorials/pdb/index.html.
These coordinates can also be found on the book website.

2.6. Atomic-Level representations of biological molecules.

(a) The following are the PDB accession number of the molecules used in this
problem or the link to the relevant websites. All of them are also available on
the book website.

• ATP: ATP.pdb from http://xray.bmc.uu.se/hicup.

• Phosphatidylcholine: stearyl-oleyl-phophacholine.pdb from

http://faculty.gvsu.edu/carlsont/mm/lipids/pg.html.

• B-DNA: bdna.pdb from http://chemistry.gsu.edu/glactone/PDB/pdb.html.

• G-actin: 1J6Z.pdb from the PDB.

• Lambda repressor/DNA complex: 1LMB.pdb from the PDB has the DNA-
binding region of lambda repressor complexed with DNA.

• Lac repressor/DNA complex: 1lbg.PDB and 1tfl.pdb from the PDB.

• Hemoglobin: 1hga.pdb from the PDB.

7
8 CHAPTER 2. RULERS AT MANY DIFFERENT SCALES

• Myoglobin: 1mbo.pdb from the PDB.

• HIV gp120: 1GC1.PDB from the PDB.
• Green fluorescent protein (GFP): 1GFL.pdb from the PDB.
• RNA polymerase: 1L9U.pdb from the PDB.

2.7. Coin flips and partitioning of fluorescent proteins.

(a) You can calculate both hn21 i and hn1 i2 by using the observation that
N
X N!
hnb1 i = nb1 pn1 q N −n1 . (2.1)
n1 =0
n1 !(N − n1 )!

If we recall that
N
X N!
(p + q)N = pn1 q N −n1 , (2.2)
n1 =0
n 1 !(N − n 1 )!

then we can find hn1 i, for example, as

N
∂ X N!
hn1 i = p pn1 q N −n1 . (2.3)
∂p n =0 n1 !(N − n1 )!
1

This same strategy can be used to find hn21 i.

(b) The key assumption for this part of the problem is I = αN , I1 = αN1 , etc.
By substituting these values into the expression for h(I1 − I2 )2 i we can beat this
into a form where we need nothing more than to just compute quantities of the
form hnb1 i as we did in part (a) of the problem.
Chapter 4

Who: Bless the Little

Beasties

4.1. Structure of hemoglobin and myoglobin.

Check out the hints for problem 2.6 for information about the relevant PDB
files. They can also be found on the book website.

4.4. Mutation correlation and physical proximity on the gene.

The key concept in the problem is that the frequency of recombination between
two alleles is proportional to the distance between them. The first thing to do
is to evaluate the ratios in table 4.2, thus obtaining “distances” whose units
Sturtevant later dubbed “centimorgans”—one centimorgan or cM corresponds
to 1% recombination frequency. Now the problem becomes one of finding the
linear ordering of the factors B, C, O, P, R, and M consistent with the distances
derived from the data.

9
Chapter 5

Mechanical and Chemical

Equilibrium in the Living
Cell

5.1. Energy cost of macromolecular synthesis.

As an example we will determine the biosynthetic cost of the amino acid valine.
We start by accessing “valine biosynthesis” in “ecocyc.org”. Clicking on “More
Detail” once shows the structures of the intermediate species.
Notice that 2 pyruvate molecules are the precursor for this amino acid. Let’s
first calculate its energy cost. In order to do that we can look at figs. 5.2Con-
structing the Cell: Managing the Mass and Energy Budget of the Cellfigure.5.2
and 5.5Constructing the Cell: Managing the Mass and Energy Budget of the
Cellfigure.5.5. In “ecocyc.org” this information can be found by searching for
“glycolysis I”.
Let’s go to “glycolysis I” and click on “More Detail”. Remember that
in order to go from glucose to glucose-6-phosphate one ATP molecules had
to be turned into ADP. This corresponds to step one in fig. 5.2Constructing
the Cell: Managing the Mass and Energy Budget of the Cellfigure.5.2. Now,
from glucose-6-phosphate we go to fructore-6-phosphate reversibly (step 2 in
fig. 5.2Constructing the Cell: Managing the Mass and Energy Budget of the
Cellfigure.5.2). Notice the double arrow between these two molecules in “gly-
colysis I”. In order to continue into fructose-1,6-biphosphate we pay one more
ATP molecules (step 3 in fig. 5.2Constructing the Cell: Managing the Mass
and Energy Budget of the Cellfigure.5.2). In step 5 we go to two molecules of
glyceraldehyde-3-phosphate reversibly.
From this point on we obtain one NADH molecule, two ATP molecules and
one pyruvate molecules per glyceraldehyde-3-phosphate molecule. In conclusion,
we started from one glucose molecule and ended up with two pyruvate molecules.

11
12 CHAPTER 5. MECHANICAL AND CHEMICAL EQUILIBRIUM

The total ATP cost is

ATP cost = +1 (5.1)

|{z}
glucose→glucose-6-phosphate
+1
|{z}
fructose-6-phosphate→fructose-1,6-phosphate
+2 × ( −1
|{z}
fructose-6-phosphate→1,3-bisphosphoglycerate
−1 ) = −2,
|{z}
phosphoenolpyruvate→pyruvate

which is a net ATP gain of one per pyruvate molecule. We also gained one
NADH molecule per pyruvate in the conversion from glyceraldehyde-3-phosphate
into 1,3-bisphosphoglycerate.
Finally, we turn to the synthesis on valine itself. We see that we start off
with two pyruvate molecules and that in the process of getting to valine we
had to pay one NADPH molecule. We also had to transform glutamate into
2-ketoglutarate. The cell will have to pay an extra energy cost to turn this
2-ketoglutarate back into glutamate.
Let’s go to “glutamate biosynthesis III”. Notice that the precursor for glu-
tamate is in fact 2-ketoglutarate. In order to obtain it one NADPH molecule
and an ammonia molecule have to be paid. Inorganic ions such as phosphate
and ammonium are present in bacterial growth medium, which always contains
a mixture of salts as well as a carbon source such as glucose. As noted in ta-
ble 5.1Constructing the Cell: Managing the Mass and Energy Budget of the
Celltable.5.1, however, we ignore the cost of phosphate and ammonium.
The net cost of making valine out of two pyruvate molecules is two NADPH
molecules. If we add the cost of making two pyruvate molecules it results in
• One glucose molecule consumed.
• Two ATP molecules gained.

• Two NADH molecules gained and two NDAPH molecules consumed.

Which means that the net cost of making a valine molecule is -2 ATP molecules
or −40 kB T (energy is gained).
Chapter 6

Entropy Rules!

6.3. Statistical mechanics of an optical trap

In this problem, we are asked to find the mean excursion hx2 i of the bead in
the trap. For simplicity, we are pretending that the bead can move in only one
dimension and moves on the energy landscape U (x) = 21 kx2 . To find
Paverages in
statistical mechanics, we follow the general prescription hf (x)i = x f (x)p(x),
where we are being intentionally vague about what x might be. For the case of
interest here, the key point is to evaluate
1 ∞ 2 −β 1 kx2
Z
2
hx i = x e 2 dx, (6.1)
Z ∞
where Z ∞
1 2
Z= e−β 2 kx dx. (6.2)
∞
Hence, the reader is really asked to perform two tasks in this problem. First,
to justify the claims above more carefully and second, to perform the relevant
calculations to actually find how the stiffness depends upon hx2 i. Note that in
our definition of Z, we have been sloppy in the sense that there are prefactors
we have ignored since they are not relevant to the actual solution of the problem
as a result of the fact that they drop out of the problem.

6.4. Free vs. bound ligand

We start by considering the different species involved in a reaction of the form
L+R LR. (6.3)
The total concentrations of each of these species can be written in terms of the
free concentrations as
Rtot = R + LR (6.4)
Ltot = L + LR (6.5)
L·R
Kd = . (6.6)
LR
(6.7)

13
14 CHAPTER 6. ENTROPY RULES!

As we have seen repeatedly, the probability of finding a ligand bound to a

receptor is given by
LR 1 1
pbound = = = . (6.8)
R + LR R/LR + 1 Kd /L + 1

From eqn. 6.8 we see that we need L in terms of Ltot , Rtot , and Kd . We do this
by using eqns. 6.4 and 6.5 to rewrite eqn. 6.6

LR = Ltot − L
R = Rtot − LR = Rtot − Ltot + L
L · (Rtot − Ltot + L)
Kd = , (6.9)
Ltot − L
which we can solve for L which can then be substituted backed into the expres-
sion for pbound .

6.9. Lattice model of the chemical potential In this problem, we are

interested in making a toy model of a solution so that we can evaluate the free
energy difference in the system between the state where there are Ns solutes
and the state where there are Ns + 1 solutes. If we imagine a lattice model for
the solution with a total Ω sites on the lattice, then the free energy of our Ns
solutes will be
Ω!
G(Ns ) = Ns s − kB T ln , (6.10)
Ns !(Ω − Ns )!
where the first term tells us the enthalpy cost associated with putting a solute
molecule in solution and the second term is the entropy associated with all the
different ways those solute molecules can be distributed in the solution. Now
to finish the problem requires two steps: 1) use the Stirling approximation to
simplify the expression for the entropy, 2) take the difference in the free energy
for the states in which there are Ns and Ns + 1 solutes. An alternative scheme is
to use the expression given above and to compute the difference directly and use
the standard algebraic rules for logarithms to simplify the resulting expression.
Chapter 7

Two-State Systems: From

Ion Channels to
Cooperative Binding

7.5. Carbon monoxide and hemoglobin.

The general expression appropriate for carrying out the numerical part of the
problem is given by
 nO2
[O2 ]
O
Kd 2
pO2 =   nO2  nCO , (7.1)
[O2 ] [CO]
1+ O + KdCO
Kd 2

with a similar expression for the binding of CO. It is worth exploring how ex-
pressions like this arise by using simple lattice models like those explored in the
chapter.

15
Chapter 8

Random Walks and the

Structure of the
Macromolecules of the Cell

8.7. Chromosome tethering.

We need to solve the partial differential equation

∂G b2 1 ∂ ∂G 1 ∂2G ∂2G
= [ (r )+ 2 + ], (8.1)
∂N 6 r ∂r ∂r r ∂θ2 ∂z 2
where G is a function of the arguments r, θ, z and N . The strategy the reader
should adopt is separation of variables by adopting a trial solution of the form

G(r, θ, z, N ) = R(r)Θ(θ)Z(z)n(N ). (8.2)

In light of this trial solution, we have

1 dn b2 1 1 ∂ ∂R 1 1 ∂2Θ 1 ∂2Z
= [ (r )+ 2 2
+ ] (8.3)
N dN 6 R r ∂r ∂r r Θ ∂θ Z ∂z 2
The concept with separation of variables is that the two sides are functions of
different variables but are equal to each other. The only way to guarantee that
is true is that both sides are equal to the same constant. For example, for the
case shown above, we have both equal to a separation constant −k 2 , resulting
in
dn
= −k 2 N, (8.4)
dN
which implies in turn
2
n(N ) = e−k N
. (8.5)

17
18 CHAPTER 8. THE STRUCTURE OF THE MACROMOLECULES

We can once again banish different terms to different sides of the equation
resulting in
b2 1 1 ∂ ∂R 1 1 ∂2Θ b2 1 ∂ 2 Z
− [ (r )+ 2 2
− k2 = . (8.6)
6 R r ∂r ∂r r Θ ∂θ 6 Z ∂z 2
Once again, given that we have a function of r and θ on the left side and a
function of z on the right side, both sides must be equal to a constant which
this time we christen −m2 . This process is repeated until we have solved four
separate differential equations for the four unknown functions. Next, the reader
needs to impose boundary conditions which amounts to making sure that the
solution vanishes on the surface of the cylinder. This leads us to the ability to
write the solution as
∞ X
∞ X
∞ α2 2
X pj b 2 2 2
+ q 6L
π b αpj
G(r, θ, z, N ) = e−( 6a2 2 )N
Jp ( r)
p=0 j=1 q=1
a
qπz
×sin (Apjq cos pθ + Bpjq sin pθ). (8.7)
L
To determine the constants Apjq and Bpjq , we now need to appeal to our “initial
condition” which tells us the origin of the polymer chain (i.e. the tether point).
In particular, we have
∞ X ∞
∞ X
X αpj qπz
G(r, θ, z, 0) = Jp ( r)sin (Apjq cos pθ + Bpjq sin pθ). (8.8)
p=0 j=1 q=1
a L

To determine the coefficients, we resort to the use of the orthonormality of the

basis functions by writing
Z 2π Z L Z a
q 0 πz αp0 j 0
G(r, θ, z, 0)sin Jp0 ( r)cos p0 θ rdrdzdθ = (8.9)
0 0 0 L a
∞ X ∞ X ∞ Z a
X αp0 j 0 αpj
= Apjq × Jp0 ( r)Jp ( r)rdr
p=0 j=1 q=1 0 a a
Z L Z 2π
qπz q 0 πz
× sin sin dz cos pθ cos p0 θdθ.
0 L L 0

The integrals on the right are all straightforward since each of them reflects a
convenient orthogonality relation of the form
Z a
αp0 j 0 αpj a2
Jp0 ( r)Jp ( r)rdr = δpp0 δjj 0 Jp+1 (αpj )2 , (8.10)
0 a a 2
and
L
q 0 πz
Z
qπz L
sin sin dz = δqq0 (8.11)
0 L L 2
and Z 2π
cos pθ cos p0 θdθ = δpp0 π. (8.12)
0
HINTS TO THE PROBLEMS 19

As was done in the chapter, to get the probability we need

G(r, θ, z, N )
p(r, θ, z, N ) = R a R L R 2π . (8.13)
0 0 0
G(r, θ, z, N ) rdrdθdz

Once the reader does this algebra, the remaining step for making contact with
the data is to figure out how to collapse the probability distribution so it is only
a function of z or the perpendicular distance from the axis of the cell.

8.8. Three-dimensional random walk and polymer cyclization.

To get the cyclization probability start by expressing the number of walks that
return to the origin using the Krönecker delta function δl as
X
Ω◦ = δ(l1 + l2 + l3 + . . . + lN ). (8.14)
{li }

The sum goes over all possible choices, which for a three-dimensional walk are:
li ∈ {±x̂, ±ŷ, ±ẑ} for each of the N steps, i = 1, 2, . . . , N . The function δl
returns 1 if l is zero, otherwise the value of the function is 0.
The next step is to use the mathematical identity
Z 2π Z 2π Z 2π
1 1 1
δ(l) = eilx kx dkx · eily ky dky · eilz kz dkz , (8.15)
2π 0 2π 0 2π 0

where the vector l = (lx , ly , lz ) has integer coordinates. This will lead to an
expression for the number of random walks that return to the origin written as
an integral from which the scaling with N can be derived.

8.12. Scaling of Protein Size

The PDB files with the crystal structures of myoglobin (1MBO), hemoglobin
(1HGA), bovine pancreatic trypsin inhibitor (1PIT), lysozyme (132L), cyto-
chrome C (3CYT), G-actin (1ATN), and an α-beta tubulin dimer (1TUB) can
be found on the book website.
Chapter 9

Electrostatics for Salty

Solutions

9.5. A simple model for viral capsid assembly.

To compute the entropy for the n-mers use a lattice model of an ideal solution
which consists of Ωn = V /vn boxes among which the Nn n-mers can be dis-
tributed. Combine this entropy with the n-mer energy to obtain the free energy
for part (a).

9.11. Toy model of salt dependent ligand-receptor binding.

(a) In order to obtain the free energies of the bound and unbound states use
the chemical potential for the ligands and the ions, assuming that they make up
an ideal gas. Since the free energy cost of removing any one of these molecular
species from solution is given by its chemical potential, the concentration depen-
dence of the chemical potential then reveals the dependence of the equilibrium
constant on salt concentration.
(b) In this case make use of the free energy of a charged ball in a salty solution
to account for the free energy of the unbound state of the ligand and receptor
pair.

21
Chapter 10

Beam Theory: Architecture

for Cells and Skeletons

10.1. Persistence length and Fourier analysis.

The Fourier expansion of the angle θ(s), that takes into account of the boundary
conditions, can be written as
∞  
X (2n − 1)πs
θ(s) = θ̃2n−1 sin . (10.1)
n=1
2L

Putting this into the equation for the beam energy leads to an expression for
the beam energy in terms of the amplitudes θ̃2n−1 . The average values of the
amplitudes can in turn be obtained from equipartition. The idea here is that the
energy of the beam acquires the form of a sum over independent modes (each
one having an energy quadratic in the amplitude), and in thermal equilibrium
each mode has an average energy of kB T .

23
Chapter 11

Biological Membranes:
From the Golgi Apparatus
to Membrane Fusion

11.7. Bending modulus and the pippete aspiration experiment.

It is much more convenient to carry out our analysis in q-space. Our Fourier
transform convention is
Z
A
h(x) = h̃(q)e−iq·x d2 q. (11.1)
(2π)2
The goal is to write the energy in Fourier transformed variables as
 2 Z
1 A
|h̃(q|2 κb q 4 + σq 2 d2 q


Etot = (11.2)
2 2π
This follows by using the real space energy and Fourier transforming. To do so,
you will need to use the tricks for Fourier transforming terms with derivatives
which can be obtained by integrating by parts. The next step is to use the
equipartition theorem to show
kB T
hh(q)2 i = (11.3)
A(κb q 4 + σq 2 )
Using the above results we can write the average difference betwen real and
projected area as:

1 A2
Z
hAact − Ai = q 2 h|h̃(q|2 id2 q (11.4)
2 (2π)2
By substituting the result for hh(q)2 i, we can then find a relation between the
tension and the area change.

25
26 CHAPTER 11. BIOLOGICAL MEMBRANES

11.8. Variational approach to deformation induced by MscL.

The first point is to obtain the partial differential equation describing the defor-
mation field. From the chapter, we know that the free energy associated with
the hydrophobic mismatch is of the form
Z Z  2
Kb Kt u
Ghydrophobic = (∇2 u)2 d2 r + d2 r.
2 2 a
In order to determine the Euler-Lagrange equations, we need to evaluate

δG[u(r)] Ghydrophobic (u(r) + η(r)) − Ghydrophobic (u(r))

= lim→0 = 0.
δu(r) 
(11.5)
What this means precisely is that the reader should literally take the free energy
expression given in eqn. 11.5 and evaluate the difference called for in eqn. 11.5.
There are many separate terms to consider and the objective is to beat them
all into the form Z ∞
δG[u(r)]
= η × stuff dr = 0. (11.6)
δu(r) R

Effectively, the differential equation of interest will be the part we have labeled
“stuff” above.
The differential equation can be shown to be related to a Bessel differential
equation. The most intuitive idea is to recognize that problems with cylindrical
symmetry like that of our ion channel often lead to equations that can be beaten
into the form of Bessel’s differential equation which is of the form

d2 u du
r2 +r + k 2 r2 u = 0, (11.7)
dr2 dr
or to the modified Bessel equation which can be written as

d2 u du
r2 +r − k 2 r2 u = 0. (11.8)
dr2 dr
Note that these two equations are not the most general form of the Bessel or
modified Bessel equation, but suffice for our purposes. The details of how to
solve such an equation itself are beyond the scope of this book, but we will
show the relation between the differential equation we need to solve given in
eqn. 11.172 and the Bessel equation. In particular, we can write the modified
Bessel equation as
∇2 u = k 2 u. (11.9)
If we now apply ∇2 to both sides of this equation, we are left with

∇4 u = k 2 ∇2 u, (11.10)

but because of the previous equation ∇2 u = k 2 u, this can be rewritten as

∇4 u = k 4 u, (11.11)
HINTS TO THE PROBLEMS 27

which is precisely our eqn. 11.172 as long as we make the identification

Kt 1
k = (− )4 (11.12)
Kb a2
As appeared in the chapter itself, we have four allowed values of k corresponding
to the four fourth roots of −1 and given by
q q √ √ 
Kt
k1 = 4
Kb a2 e
iπ/4
= 4 KKb at 2 22 + i 22 (11.13)
q q  √ √ 
k2 = 4 KKb at 2 ei3π/4 = 4 KKb at 2 − 22 + i 22 (11.14)
q q  √ √ 
k3 = 4 KKb at 2 ei5π/4 = 4 KKb at 2 − 22 − i 22 (11.15)
q q √ √ 
Kt
k4 = 4
Kb a2 e
i7π/4
= 4 KKb at 2 22 − i 22 . (11.16)

The solutions to our differential equation can be written down directly as

modified Bessel functions of the form K0 (kr) and I0 (kr), though from the outset
we can reject the solutions I0 (kr) since at large r they diverge, inconsistent with
our boundary conditions that u(∞) = 0 and u0 (∞) = 0. Further, K0 (k2 r) and
K0 (k3 r) diverge for large r. As a result, we write our solution to the problem
as
u(r) = A1 K0 (k1 r) + A4 K0 (k1∗ r) (11.17)
To determine the coefficients A1 and A4 , we need to apply our boundary con-
ditions u(R) = U0 and u0 (R) = 0. The displacement boundary condition at the
protein edge results in

u(R) = A1 K0 (k1 R) + A4 K0 (k1∗ R) = U0 . (11.18)

Similarly, we can evaluate the slope at the protein boundary resulting in

dK0 (k1 r) dK0 (k1∗ r)
u0 (R) = A1 + A4 = 0. (11.19)
dr dr
Using the fact that K00 (x) = −K1 (x), we can rewrite this boundary condition
in the more convenient form

u0 (R) = −A1 k1 K1 (k1 R) − A4 k1∗ K1 (k1∗ R) = 0. (11.20)

We can now solve these two equations for A1 and A4 resulting in

k1 K1 (k1 R)
A1 = U0 , (11.21)
k1∗ K0 (k1 R)K1 (k1∗ R) − k1 K0 (k1∗ R)K1 (k1 R)
and
k1∗ K1 (k1∗ R)
A4 = −U0 . (11.22)
k1∗ K0 (k1 R)K1 (k1∗ R) − k1 K0 (k1∗ R)K1 (k1 R)
Using these values for A1 and A4 , we can now plug them back into eqn. 11.17
to yield the deformation profile around the ion channel.
28 CHAPTER 11. BIOLOGICAL MEMBRANES

One way to obtain the energy is to use the divergence theorem which permits
us to convert the energy over the entire membrane into an integral just around
the boundary of the protein itself. Effectively, the divergence theorem in this
form amounts to a sophisticated version of integration by parts. The theorem
can be stated as Z Z
F · n ds = ∇ · F dA. (11.23)
∂Ω Ω
Our aim is to figure out how the free energy of deformation can be written such
that it is clear that the right side is the divergence of some vector. In particular,
our goal is to rewrite the term (∇2 u)2 which can be built around the observation
that
∇ · (∇2 u∇u) = ∇3 u · ∇u + (∇2 u)2 . (11.24)
In addition, we have

∇ · (u∇3 u) = ∇3 u · ∇u + u∇4 u, (11.25)

which means that if we subtract these two equations, we can isolate (∇2 u)2 as

(∇2 u)2 = −∇ · (u∇3 u − ∇2 u∇u) + u∇4 u. (11.26)

We can now plug this way of rewriting (∇2 u)2 back into the expression for the
free energy of deformation. The result is
Z Z  2
Kb Kt u
Ghydrophobic = [−∇ · (u∇3 u − ∇2 u∇u) + u∇4 u]d2 r + d2 r.
2 2 a
This very interesting result brings us much closer to having an expression that
will permit us to invoke the divergence theorem. In particular, if we rewrite this
as
Z Z
Kb Kt Kb 4
Ghydrophobic = [−∇ · (u∇3 u − ∇2 u∇u)]d2 r + ( 2 u + ∇ u)ud2 r,
2 2a 2
(11.27)
we note that the term in parentheses in the second integral is nothing more than
the equilibrium differential equation we solved in the first place and is thus zero.
Hence, we can now rewrite the energy as
Z Z
Kb 3 2 2 Kb
Ghydrophobic = [−∇·(u∇ u−∇ u∇u)]d r = − r·(u∇3 u−∇2 u∇u)rdθ.
2 2
(11.28)
Chapter 13

Random Walks: A
Statistical View of
Biological Dynamics

13.2. Diffusion from a point source.

The diffusion equation reads:

∂ ∂2
c(x, t) = D 2 c(x, t). (13.1)
∂t ∂x
Let’s √
take the Fourier transform of both sides (which just means multiplying by
eikx / 2π and integrating) and take stock of the situation:
Z +∞ Z +∞
1 ∂ D ∂2
√ eikx c(x, t) dx = √ eikx 2 c(x, t) dx (13.2)
2π −∞ ∂t 2π −∞ ∂x
Z +∞ Z +∞
∂2
 
∂ 1 ikx D
√ e c(x, t) dx = √ eikx 2 c(x, t) dx (13.3)
∂t 2π −∞ 2π −∞ ∂x
Z +∞
∂ D ∂2
c̃(k, t) = √ eikx 2 c(x, t) dx. (13.4)
∂t 2π −∞ ∂x

All that we’ve done so far is to exchange the order of integration and differenti-
ation on the LHS and then used the definition of c̃. Now comes the important
part. The key is to integrate by parts on the RHS to get the partial deriva-
tives w.r.t x off of c(x, t) and move them onto the exponential instead. How is
this done? Recall that for two functions F and G that both depend on x the
following identity holds (the product rule):

∂ ∂F ∂G
(F G) = G+F . (13.5)
∂x ∂x ∂x

29
30 CHAPTER 13. A STATISTICAL VIEW OF BIOLOGICAL DYNAMICS

Now rearrange things and integrate:

Z +∞ Z +∞ Z +∞
∂G ∂ ∂F
F dx = (F G) dx − G dx (13.6)
−∞ ∂x −∞ ∂x −∞ ∂x
+∞ Z +∞
∂F
= (F G) − G dx. (13.7)
−∞ −∞ ∂x

Let’s suppose that either or both of F and G vanish at ±∞, then we have:
Z +∞ Z +∞
∂G ∂F
F dx = − G dx. (13.8)
−∞ ∂x −∞ ∂x

This is just what we wanted; we’ve succeeded in moving the derivative off of G
on the LHS and onto F on the RHS, at the cost of a minus sign. Let’s go back
to Eq. 13.4 and integrate by parts:
Z +∞
∂ D ∂2
c̃(k, t) = √ eikx 2 c(x, t) dx (13.9)
∂t 2π −∞ ∂x
  +∞
D ∂c
= √ eikx
2π ∂x −∞
Z +∞  
D ∂ ikx ∂
−√ e c(x, t) dx (13.10)
2π −∞ ∂x ∂x
Z +∞
D ∂
= −√ (ik)eikx c(x, t) dx (13.11)
2π −∞ ∂x
  +∞ Z +∞
Dik D
= − √ eikx c(x, t) +√ (ik) (13.12)
2π −∞ 2π −∞
 
∂ ikx
× e c(x, t) dx
∂x
Z +∞
D
= + √ (ik)2 eikx c(x, t) dx (13.13)
2π −∞
= −k 2 Dc̃(k, t). (13.14)

We integrated by parts twice and had to assume that the flux ∂c/∂x and con-
centration c(x, t) both vanish at infinity which are both very reasonable assump-
tions. At first glance it may seem like we haven’t gained that much. But hold
on! Instead of having derivatives w.r.t. both x and t, we now have derivatives
w.r.t. time t only! We have transformed our PDE into an ODE; c̃ is a function
of k but for our purposes it’s just a parameter; let’s rewrite our equation to
make this all explicit:
d
c̃(t; k) = −k 2 D c̃(t; k). (13.15)
dt
Of course this is an equation we know very well how to solve, and we can just
write down the answer: 2
c̃(k, t) = c̃(k, 0)e−k Dt . (13.16)
HINTS TO THE PROBLEMS 31

To get c̃(k, 0) we need to Fourier transform our initial condition. Once we do

that, all that remains effectively is to perform the inverse Fourier transform
using the
Z +∞
1
c(x, t) = √ e−ikx c̃(x, t) dk (13.17)
2π −∞
c0 +∞ −ikx −k2 Dt
Z
= e e dk. (13.18)
2π −∞

13.5. Two-Dimensional FRAP Analysis

In order to solve this problem we will use separation of variables. We will assume
that c(r, t) = T (t)ρ(r), which turns our diffusion equation of interest into

d2 ρ
   
1 dT 11 d dρ 1 1 dρ
= r = +r 2 . (13.19)
D T dt ρ r dr dr ρ r dr dr
Since both sides of the equation depend on different variables, they must be
equal to a constant, which we set equal to K 2 . This means that the solution
for the temporal part is
2
T (t) = e−DK t
(13.20)
and the resulting equation for the spatial part

d2 ρ dρ
r2 2
+r + K 2 r2 ρ = 0. (13.21)
dr dr
This last equation can be written in a Bessel form by making the substitution
z = kr. Once the solutions are in hand, the boundary conditions need to
be imposed. In particular, we have the following expression for the boundary
condition

J00 (kR) = 0. (13.22)

The roots of J00 are just the roots of J1 because of the identity J00 (z) = −J1 .
This will determine discrete values for Ki . The full solution is
∞
X 2
c(r, t) = a0 + ai e−DKi t J0 (Ki r). (13.23)
i=1

We can determine the coefficients ai using the initial condition c(r, 0).
Chapter 15

Rate Equations and

Dynamics in the Cell:
Cytoskeleton Under
Construction

15.6. Microtubule dynamics and dynamic instability.

To write down the master equations we have to consider all the possible tran-
sitions to and from the state of interest. For example, in the case of writing
p+ (n, t), which is the probability of finding a growing filament of length n at
time t we have the following options:

• A filament with length n − 1 which grows into a filament with length n.

This process happens with a speed v+ and its contribution to the master
equation is +v+ p+ (n − 1, t).

• The filament with length n might grow into a filament with length n + 1,
contributing with a term −v+ p+ (n, t).

• A filament that is growing can undergo a “catastrophe”, which turns it

into a shrinking one and adds the term −f+− p+ (n, t).

• Finally, a filament that was shrinking can undergo a “rescue” and turn
into a growing one, giving the term +f−+ p− (n, t).

the master equation for a growing filament is therefore:

∂p+ (n, t)
= v+ (p+ (n − 1, t) − p+ (n, t)) − f+− p+ (n, t) + f−+ p− (n, t) (15.1)
∂t

33
34 CHAPTER 15. CYTOSKELETON UNDER CONSTRUCTION

We can now do a Taylor expansion of the term with p+ (n − 1, t) ' p+ (n, t) −

∂p+ (n,t)
∂n ,
To solve the differential equations that you will obtain from the Taylor ex-
pansion strategy, assume steady state (i.e. ∂p+∂t(n,t) = ∂p−∂t(n,t) = 0. This implies
in turn that

dp+ (n, t)
v+ = f−+ p− − f+− p+ (15.2)
dn
dp− (n, t)
v− = f−+ p− − f+− p+ .
dn
This is a linear system of first order differential equations. We can write the
RHS of both equations in matrix form using
 
−f+− /v+ f−+ /v+
(15.3)
−f+− /v− f−+ /v−
We want to look for its eigenvalues and eigenvectors which will permit us to
write down the solutions themselves.

15.8. Antenna model for microtubule length control.

(a) Every monomer of the filament can be characterized by its Kip3 occupancy,
which we denote ni ; here i = 1, 2, . . . , N where N = L/a is the number of
monomers in the filament, and a is the length of a single monomer. In steady
state the number of Kip3 molecules arriving at the ith monomer per unit time
must equal the number of monomers departing the same monomer. In the model
we assume that all the monomers in front of the ith one funnel Kip3 molecules to
it (i.e. Kip3 molecules do not detach before they fall off the end of the filament)
and therefore the balance between incoming and outgoing Kip3 becomes
ikbind = vni . (15.4)
Given these arguments, one can now work out the steady-state occupancy and
how it depends upon position on the filament. As with many problems in the
book, there are multiple ways to solve the problem. An alternative approach to
that described above is to write down a differential equation of the form
dn kbind
= (15.5)
dx av
which implies that
kbind kbind
n(x) = x=i . (15.6)
av v
This differential equation arises by writing a mass balance for each monomer
and then making a continuum approximation.
(b) The flux of Kip3 into the terminal monomer in steady state is N kbind =
kbind L/a, therefore the depolymerization rate for the filament is kof f =
aN kbind = kbind L. The way to see this is to use the result of part (a) which
tells us how the number of Kip3 molecules depends upon position.
Chapter 16

Dynamics of Molecular
Motors

16.4. Kinetics of two-state motors

Start by taking the continuum limit of eqns. 16.33 and 16.34 so that differences
in the variable n become derivatives of the probabilities with respect to this
parameter. Then substitute the trial solution into the two differential equations
to obtain a two-by-two system of linear equations for C0 and C1 , where
   
p0 (n, t) C0
= ei(Kn−ωt) . (16.1)
p1 (n, t) C1

Now, if we write the equations for C0 and C1 in matrix form as

− + + −
(1 − iK − 21 K 2 ) + kB
    
iω − (kA + kB ) kA C0 0
− + + − =
kA (1 + iK − − 12 K 2 ) + kB iω − (kA + kB ) C1 0
(16.2)
then the matrix on the left-hand side must have a zero determinant for a solution
to exist.

35
Chapter 17

Biological Electricity and

the Hodgkin-Huxley Model

17.6. Time-dependent cable equation

(a) The time-dependent cable equation 17.30 is a partial differential equation.
We can solve it numerically using Matlab. Alternatively, Mathematica or any
similar program can be used.
The function pdepe solves equations of the general form
      
∂u ∂ ∂u ∂u
c x, t, u, = x−m xm f x, t, u, + s x, t, u, . (17.1)
∂x ∂x ∂x ∂x
When we compare this to eqn. 17.30 we can make the following identifications
u(x, t) = V (x, t) (17.2)
 
∂V (x, t)
c x, t, V (x, t), = −τ
∂x
 m = 0
∂V (x, t) ∂V (x, t)
f x, t, V (x, t), = −λ2
∂x ∂x
 
∂V (x, t) gN a (V (x, t))
s x, t, V (x, t), = (V (x, t) − V (x, t)N a K
N ernst ) + (V (x, t) − V (x, t)N ernst ).
∂x gK
This function is called in the following way: sol = pdepe(m,pdefun,icfun,
bcfun,xmesh,tspan), where pdefun returns the values of c, f, and s ([c,f,s]
= pdefun(x,t,u,DuDx)). The initial conditions are given by
u(x, t0 ) = u0 (x), (17.3)
using u = icfun(x). Finally, the boundary conditions correspond to the equa-
tion  
∂u
p(x, t, u) + q(x, t) f x, t, u, = 0. (17.4)
∂x

37
38 CHAPTER 17. BIOLOGICAL ELECTRICITY

We input these conditions using [pl,ql,pr,qr] = bcfun(xl,ul,xr,ur,t).

Here, l and r correspond to the values of x, u, p, and q at the left and right
boundaries, respectively.

17.7. Propagation of spikes

Like in the previous problem we can use the function pdepe form Matlab. The
biggest difference is that this time we’ll use it to solve a system of differen-
tial equations rather than only one differential equation. These equations will
be the cable equation as well as the equations describing the dynamics of the
populations of closed, open and inactive Na+ channels shown in eqn. 17.44:
∂pC
= −kopen (V )pC (17.5)
∂t
∂pO
= kopen (V )pC − kinactive (V )pO
∂t
∂pI
= kinactive (V )pO
∂t
∂V (x, t) ∂ 2 V (x, t) gN a (V )
−τ = −λ2 (V (x, t) − VNNernst
a
)
∂t ∂x2 gK
+(V (x, t) − VNKernst ).

We define different column vectors that will be used in pdepe:

u(1) = pC (V ) (17.6)
u(2) = pO (V )
u(3) = pI (V )
u(4) = V (x, t)

Each of these will have corresponding c, f and s column vectors.

You will have to implement eqns. 17.45 and 17.46 in you code. For the
boundary conditions, we still set the derivative of the voltage to be zero. We fix
the channels at the boundaries to the closed state. This will certainly introduce
artifacts if the spikes reach the boundaries in our simulation. Finally, we use
the exact same initial spike as as in problem 17.6 for our initial condition, while
we set all the channels to the closed state initially.
Chapter 18

Light and Life

18.7. Eyes and the diffraction limit

(a) Remember that the vector ~r pointing to the plane of observation is given
by
~r = xx̂ + y ŷ + Z ẑ, (18.1)
where x and y are the positions along the plane and Z is the distance from the
point source to the plane. Notice that we are assuming that Z  x, y, which
means that we can Taylor expand r accordingly.
(b) We want now to write the field at position r0 on the plane coming from the
patch dS of our source. Using the definitions from Figure 18.39(C) we can write
r0 as p
r0 = (R cos φ − r)2 + (R sin φ)2 + Z 2 . (18.2)
Note that, like in part (a), we can Taylor expand this expression to first order
in r. This will allow to obtain a simple functional for the disturbance from dS
at the plane given by
A ~
ψ = e−i(ωt−k·~r) . (18.3)
r
(c) By integrating the disturbances over the source (over R and φ) you will
reach an expression of the form
Z 2π
eia cos(φ) dφ = 2πJ0 (a), (18.4)
0

where Jn (x) is the n-th Bessel function of the first kind. Another useful identify
to keep in mind is the one relating the integral of Bessel functions such as
Z
J0 (x)x dx = xJ1 (x). (18.5)

(d) The first zero of the intensity, related to the resolution limit, will be given by
the first zero of the first Bessel function of the first kind, J1 (x). The zeros of this

39
40 CHAPTER 18. LIGHT AND LIFE

function can be found in tables of integrals or in programs such as Mathematica.

In the following table we show the first few zeros for reference.
Zero number for J1 (x) x
1 3.83171
2 7.01559
3 10.1735
Chapter 19

Organization of Biological
Networks

19.10. Copy number and the Poisson promoter

(a) f is the fraction of the cell cycle during which there is one copy of the gene.
As a result of the change in the number of copies of the gene of interest, the
mean number of mRNAs varies depending upon when in the cell cycle we look.
The hint is to note that when there are two copies of the gene of interest, then
m −2λ
p(m) has a contribution of the form (2λ)m!e . What fraction of the cell cycle
is this the relevant p(m)?
(b) We can use the result of part (a) to find the average number of mRNAs as
∞ ∞ ∞
X X λm e−λ X (2λ)m e−2λ
hmi = mp(m) = f m + (1 − f ) m . (19.1)
m=0 m=0
m! m=0
m!

We can use our usual trick of differentiation with respect to a parameter by

remembering that
∞ ∞
X λm e−λ d X λm
m = e−λ λ . (19.2)
m=0
m! dλ m=0 m!
Using this insight, the remainder of the problem comes down to algebra.
(c) Similarly, we can evaluate
∞
X
hm2 i = m2 p(m) (19.3)
m=0

using the same tricks as in part (b).

41
Chapter 21

Sequences, Specificity, and

Evolution

21.4. Nucleosome urn and chemical potentials

(a) The chemical potential of nucleosomes and base pairs on genomic DNA is
related to the free energy G(nnuc , nbp ), where nnuc is the number of nucleosomes
and nbp the number of free (i.e., not covered by nucleosomes) base pairs. In
particular, use the definitions

µnuc = G(nnuc , nbp ) − G(nnuc − 1, nbp ) (21.1)

µbp = G(nnuc , nbp ) − G(nnuc , nbp − 1) .

With these definitions, the calculations of the chemical potentials comes down
to figuring out the entropy for the number of ways of arranging the base pairs
and nucleosomes.
(b) With the chemical potentials in hand we turn to computing the probability
p(i) that the ith site of a genomic region of length N is occupied by a nucleosome,
given that the origin is nucleosome free. This probability, p(i) = 1 − p̄(i), where
p̄(i) is the probability that the ith site is nucleosome free (i.e., it is a ”base pair”
in the sense of the urn model). If we introduce
nmax
nuc
X [(i − nnuc d) + nnuc ]!
S(i) = eβ(i−nnuc d)µbp eβnnuc µnuc , (21.2)
n =0
(i − n nuc d)!n nuc !
nuc

as in the text, as the thermodynamic weight that takes into account all the
possible ways of covering i sites on the genome with base pairs and nucleosomes,
then
S(i − 1)S(N − i)
p̄(i) = . (21.3)
S(N )
The numerator in the above expression is the weight of all the states in which
the ith site is not covered by a nucleosome while the denominator is the weight

43
44 CHAPTER 21. SEQUENCES, SPECIFICITY, AND EVOLUTION

of all the states regardless of whether the ith site is nucleosome free or not.
The numerator is obtained by considering the ith site nucleosome free and then
noting that this leaves the first i − 1 sites before it and the N − i sites after it
free for occupying with nucleosomes or base pairs. Since occupying these two
stretches of genomic DNA with free base pairs and nucleosomes are independent
of each other, the total weight is the product of weights, S(i-1) S(N-i).
We can use Mathematica to make plots of p(i) for different values of the
average linker length hLi.