Why Is Mathematical

Biology So Hard?
Michael C. Reed

Although there is a long history of the applica- dogmas—such as “evolution by natural selection”,
tions of mathematics to biology, only recently has “no inheritance of acquired characteristics”, or
mathematical biology become an accepted branch “DNA → RNA → proteins”. But these are not trans-
of applied mathematics. Undergraduates are doing latable into mathematical equations or other struc-
research projects and graduate students are writ- tures without hosts of additional facts and as-
ing Ph.D. dissertations in mathematical biology, sumptions that are context-dependent. This means
and departments are trying to hire them. But what that mathematical biology is very unsatisfying for
should the Ph.D. training consist of? How should pure mathematicians, who usually are interested
departments judge work in mathematical biology? in discovering fundamental and universal structural
Such policy questions are always important and relationships. It also means that there is no “math-
controversial, but they are particularly difficult ematics of biology” in the same way that ordinary
here because mathematical biology is very differ- differential equations is the mathematics of clas-
ent from the traditional applications of mathe- sical mechanics and partial differential equations
matics in physics. I’ll begin by discussing the is the mathematics of continuum mechanics.
nature of the field itself and then return to the Diverse, yet special. Because of evolution, bio-
policy questions. logical systems are exceptionally diverse, complex,
Where’s Newton’s Law? The phenomena that and special at the same time, and this presents sev-
mathematical biology seeks to understand and pre- eral difficulties to a mathematician. The first is
dict are very rich and diverse and not derived from choosing what to work on. There’s too much biol-
a few simple principles. Consider, in comparison, ogy! How do changes in the physics or chemistry
classical mechanics and continuum mechanics. of a particular environment affect the species that
Newton’s Law of Motion is not just a central ex- live in the environment (ecology)? How do diseases
planatory principle; it also gives an immediate way spread within a population (epidemiology)? How
to write down equations governing the important do the organ systems of the human body work
variables in a real or hypothetical physical situa- (physiology)? How do the neurons in our brain
tion. Since the Navier-Stokes equations express work together to allow us to think and feel and
Newton’s Law for fluids, they are fundamental and calculate and read (neurobiology)? How does our
have embedded in them both the fundamental immune system protect us, and what are the
principle and the complexity of the fluid phe- dynamical changes that occur when we are under
nomena that we see. Thus a pure mathematician attack by pathogens (immunology)? How do cells
who proves a theorem about the Navier-Stokes use physics and chemistry to accomplish funda-
equations and an applied mathematician who mental tasks (cell biology, biochemistry)? How does
develops new numerical tools knows that he or she the genetic code, inscribed in a cell’s DNA, give
has really contributed something. Alas, there are rise to a cell’s biochemical functioning (molecular
no such central fundamental principles in biology. biology and biochemistry)? How do DNA sequences
There are principles of course—some would say evolve due to environmental pressures and random
events (genomics and genetics)?
Michael C. Reed is professor of mathematics at Duke The second difficulty is that a priori reasoning
University. His email address is reed@math.duke.edu. is frequently misleading. By “a priori reasoning” I

338 NOTICES OF THE AMS VOLUME 51, NUMBER 3

mean thinking how we would design a mechanism groups of cells in the immune system give rise to
to accomplish a particular task. As a simple the overall immune response? How do the proper-
example, both birds and planes burn something ties of individual bees give rise to the behavior of
to create energy that can be turned into potential the hive? How do the cells in a leaf “cooperate” to
energy, and both have to use the properties of turn the leaf towards the sun? How does the
fluids that are implicit in the Navier-Stokes equa- varied behavior of individuals contribute to the
tions. But that doesn’t mean that one understands spread of epidemics?
birds if one understands planes. To understand how We are familiar with these types of questions
a bird flies, one has to study the bird. Modelers are from physics. What are the right variables to
sometimes satisfied that they have created a describe the behavior of a gas, and how do the
mathematical model that “captures” the biological values of these variables arise from the classical
behavior. But that is not enough. Our purpose is mechanics of the molecules making up the gas?
to understand how the biological mechanisms give The behavior at the higher level is relatively sim-
rise to the biological behavior. Since these biolog- ple, and Newton’s law suggests the few important
ical mechanisms have been “designed” by evolution, variables at the lower level; even so, the proofs are
they are often complicated, subtle, and very spe- not easy. In the case of biological systems these
cial or unusual. To understand them, one must questions are even more difficult, because the
immerse oneself in the messy, complex details of objects at the lower level have been designed by
the biology; that is, you must work directly with evolution (or trained by feedback control; see below)
biologists. to have just the right special properties to give
Thirdly, different species (or different tissues or rise to the (often complicated) behavior at the
different cells) may accomplish the same task by higher level. And it is usually not easy to decide
different mechanisms. An astounding array of spe- what the important variables are at the lower level.
cial mechanisms allows animals to exploit special If your model has too few, you will not be study-
niches in their environments. For example, a diverse ing the “real” biological mechanism. If your model
set of locomotory mechanisms are used at differ- has too many, it may be so complicated that a
ent size scales. Thus, when you have understood lifetime of computer simulations will not give new
bird flight completely, you have not even started biological understanding. You need ideas, guess-
on the butterfly or the fruit fly. So, even when one work, experience, and luck. You need to be able to
is successful, one may have provided understand- deduce the consequences from the assumptions.
ing only in particular cases. That is what mathematicians are good at.
We can already draw some conclusions. Don’t do The difficulty of experimentation. We mathe-
mathematical biology to satisfy a desire to find uni- maticians often have an overly simple view of ex-
versal structural relationships; you’ll be disap- periments and the role they play, probably because
pointed. Don’t waste time developing “methods of we don’t conduct them ourselves. A theory is tested
mathematical biology”; the problems are too by deciding on a few crucial variables and design-
diverse for central methods. What’s left is the bi- ing the right experimental setup. For example, one
ology. You should do mathematical biology only if measures how fast metal beads and feathers fall
you are deeply interested in the science itself. If you in a vacuum or the angle subtended by two stars.
are, there’s lots of good news. We mathematicians However, the complicated histories of interaction
are experts at thinking through complex relation- between theory and experiment in quantum me-
ships and formulating scientific questions as math- chanics, nuclear physics, and elementary particle
ematical questions. Some of these mathematical physics in the twentieth century show that this
questions are deep and interesting problems in simple view is naive. And for several reasons the
pure mathematics. And most biologists know that experimental situation is even more difficult in
the scientific questions are difficult and compli- biology.
cated, so they want our help. There’s some bad news First, one is often interested in how the behav-
too; there are three more reasons why the field is ior at one level arises from lower levels. Typically,
so hard. this “emergent” behavior cannot be seen in any of
The problem of levels. In many biological prob- the parts at the lower level but arises because of
lems one is trying to understand how the behavior complex interactions among the parts. Unfortu-
of the system at one level arises from structures and nately, it can be misleading to study the parts in
mechanisms at lower levels. How does the coordi- isolation. For example, I try to understand how
nated firing of neurons give rise to the graceful certain biochemical networks in mammalian cells
motion of an arm? How does the genetic code in function. The networks give rise to systems of
DNA create, maintain, and adjust a cell’s bio- ODEs in which the nonlinear terms depend on the
chemistry? How does the biochemistry of a cell enzyme kinetics for each separate enzyme. The
allow it to receive signals, process them, and send enzymes can be isolated and their reaction kinet-
signals to other cells? How does the behavior of ics studied “in vitro” in experiments that combine

MARCH 2004 NOTICES OF THE AMS 339

 pure enzyme with pure substrate. But in the soup, However, most are very stable, and it is almost a
in the real cell, the enzymes and substrates are bind- tautology to say so, because they must all operate
ing to or being affected by other chemicals too, so in the face of changing and fluctuating environ-
one is unsure whether the “in vitro” experiments mental parameters; so if they weren’t stable, they
reflect the true “in vivo” kinetics. That is, each of wouldn’t be here. We are familiar from engineer-
the parts at the lower level behaves differently in ing with the concept of feedback control, whereby
isolation than when it is connected to the other variables are sensed and parameters are then reset
parts. to change the behavior of the system. The nephrons
Second, chance plays a role, not only in experi- in the kidney sense NaCl concentration in the blood
mentation, but perhaps in explanation. Why are and adjust filtration rate to regulate salt and water
two neighboring fields dominated by black-eyed balance in the body. The baroreceptor loop regu-
Susans and poppies respectively? Are the soils lates blood pressure, heart rate, and peripheral
different? Are the local plant and animal species resistance to adjust the circulation to different
different? Or perhaps the “explanation” is a chance challenges. Numerous such control systems are
event in the past (or all of the above). known and studied in animal and plant physiology.
Third, individuals (whether cells or flowers or There is another kind of feedback that operates
people) are both similar and different. How does between levels that poses special problems. Here
one know whether data collected is “special” or are two examples. In the auditory system sensory
“typical”? How does one assure oneself that rat data information is transformed in the cochlea to elec-
tells us something about humans or, indeed, some- trical information that proceeds up the VIIIth nerve
thing about other rats? to the cochlear nucleus and from there to various
Finally, it is characteristic of living systems that other nuclei in the brain stem (a nucleus is a large
the parts themselves are not fixed but ever chang- anatomically distinct group of cells) and on to the
ing, sometimes even affected by the behavior of the midbrain and the cortex. Surprisingly, there are also
whole (see feedback control, below). A simple true neural projections from the cortex that influence
story illustrates this point. An experimenter (#1) the sensitivity of the cochlea. Second, the dogma
who used rats in his experiments was getting very DNA → RNA → proteins → function has turned
unusual results, and the results were not repeat- out to be a naive fiction. Genes (pieces of DNA) don’t
able week to week. After six months of this he had turn themselves on or off but are activated or in-
to stop his experiments and investigate the rats. hibited by proteins. That is, proteins affect the
His rats were housed in his university’s vivarium. genes, adding a reverse loop to the simple picture,
It turned out that another experimenter (#2) had implying of course that the genes affect each other
a mean technician, and when the other experi- through the proteins. The “fundamental” objects,
menter’s technician came to get #2’s rats, #2’s rats that is, the objects most closely related to “func-
would cry out, upsetting the rats belonging to #1. tion”, may not be genes or proteins but small
The behavior of #1’s rats in experiments depended networks involving both genes and proteins that
on whether #2 was doing an experiment the same respond in certain ways to changes in the cell’s
day! Whew, I’m glad I became a mathematician. environment. These kinds of examples show that
For all these reasons, biological data must be the nineteenth-century picture of a machine with
approached cautiously and critically. Since bio- parts is a very inappropriate metaphor for (at least
logical systems are so diverse and everything some) biological systems. When there is feedback
seems to interact with everything else, there are between levels, it is hard to say which are the parts!
many possible measurements, and enormous In fact, it may be hard to say which are the levels,
amounts of data can be produced. But data itself and therefore our traditional scientific research
is not understanding. Understanding requires a paradigm of breaking things into smaller parts
conceptual framework (that is, a theory) that (lower levels) may not be successful.
identifies the fundamental variables and their This is not just a philosophical point but a fun-
causative influences on each other. In messy bio- damental research issue that deepens the impact
logical problems, without simple fundamental of the previous four difficulties. Take, for example,
principles like Newton’s Law, useful conceptual the question of dendritic geometry. It’s been one
frameworks are not easy to propose or validate. One hundred years since Ramon y Cajal made beauti-
must have ideas about how the structure of (or ful drawings of complicated dendritic arbors on
behavior of) the whole is related to the assumptions nerve cells. Is the geometry important? Surely it
about the parts. Thinking through such ideas must be, we feel, since cells in the same brain nu-
and proving the consequences of the assumptions cleus in different individuals seem to have roughly
are important ways that we mathematicians can similar dendritic arborization. And, indeed, there
make contributions. are examples where it is understood how specific
The problem of feedback control. It is com- dendritic geometry creates specific neuron-firing
mon to think of biological systems as fragile. properties and presumably specific cell function,

340 NOTICES OF THE AMS VOLUME 51, NUMBER 3

though it is not always clear what “function” means Graduate education. There is quite a bit of dis-
for a single cell embedded in layers and layers of agreement about the proper mathematics graduate
a large neural network. On the other hand, suppose training of a student who wants to be a mathe-
a cell is part of a large neural network whose matical biologist. I’ll simplify the discussion into two
job it is to transform its pattern of inputs into a (extreme) positions. The first position emphasizes
corresponding pattern of outputs. This neural maximal contact with biology and biologists as
network may have been trained to do this job by part of graduate training. Graduate students should
feedback control from a higher level, in which case take biology courses (including labs) and should par-
the details of the dendritic geometry (and even the ticipate in or even initiate collaborative modeling
details of the neural connections) may not be im- projects. This way they learn a lot of biology, and,
portant at all. The details arose from the training, even more importantly, they learn modeling and
and they are whatever they need to be to give the how to communicate with biologists. By doing this
behavior at the higher level. Furthermore, for large they study less mathematics of course, but they can
networks there may be many choices of details learn what they need later when they need it. The
that give the same network behavior, in which case second position emphasizes training as a mathe-
it will be hard to infer the behavior of the whole matician first. Graduate students should receive the
by studying the properties of the parts. traditional training in analysis or applied mathe-
I now want to turn to the policy questions that matics (or other subjects), and (ideally) the thesis
I mentioned at the beginning. I have been using should contain some applications to biological
the term “mathematical biology” to refer in the problems. But the graduate student should not
broadest way to quantitative methods in the bio- spend too much time slogging around in the bio-
logical and medical sciences. Physicists, chemists, logical details or working on collaborative projects.
computer scientists, and biological and medical It is the job of the thesis advisor to be the interface
researchers with some mathematical training can between the graduate student and the biology.
Later, after the Ph.D., when the mathematician is es-
and do contribute to the field I have been referring
tablished, he or she can choose to become involved
to as “mathematical biology”. But let us now narrow
in collaborations and learn more biology.
the focus to mathematics education, both under-
I guess that most mathematical biologists sup-
graduate and graduate, and the mathematics job
port the first position. I support the second,
market.
perhaps because that is the route that I followed
Undergraduate education. Mathematical biology
myself. Mathematical biology is really a very hard
is an extremely appealing subject to undergradu-
subject (I hope I have convinced you of that), and
ate students with good training in freshman and
a great many ideas and techniques from different
sophomore mathematics. Many are naturally in-
branches of mathematics have proven useful. So
terested in biology, and all know that we are in the
mathematical biologists need broad training in
midst of a revolution in the biological sciences. mathematics. Secondly, I believe that only deep
They are usually amazed and delighted that the and rigorous graduate training creates mathe-
mathematical techniques that they have learned can maticians who can not only learn new mathemat-
be used to help understand how biological sys- ics when they need it, but who can also recognize
tems work. Further, mathematical biology is a per- what they need to learn.
fect subject for undergraduate research projects. Hiring issues. Hiring a mathematical biologist
Biology is so diverse and so little quantitative mod- poses special challenges for departments. Most math-
eling has been done that it is relatively easy to find ematicians have no idea how large the field “biology”
projects that use undergraduate mathematics in is or how large the research communities are. Two
new biological applications. The students find such examples illustrate this. Here at Duke, Arts and Sci-
projects to be very rewarding. They know that the ences has 469 tenure-track faculty members, and
undergraduate major consists mostly of nineteenth- the Medical Center has 767 (not counting clinical
century mathematics; of course they are excited by faculty). My colleague Harold Layton is a mathe-
twenty-first-century applications. Here at Duke we matician who works on the kidney, so of course he
have found that the availability of projects in goes to the annual meeting of the American Society
mathematical biology has attracted many students for Nephrology, where the typical registration num-
to the mathematics major. Of course, it helps to ber, 12,000, completely dwarfs the registration at
have a mathematical biologist on the faculty, but the Joint Mathematics Meetings. And those consti-
it is not necessary. Any mathematician can create tute just a subset of the researchers who work on the
and supervise such projects by working coopera- kidney! So, the first issue is thinking about what kind
tively with local biologists. It requires only the of mathematical biologist you want. It is a good idea
effort to make the connections and tolerance for to involve local biologists and medical researchers
appearing “nonexpert” to the students (something in preliminary discussions, both to educate depart-
we are not used to!). ment faculty and to understand the local context.

MARCH 2004 NOTICES OF THE AMS 341

 The question of how best to judge job candidates
is particularly difficult in mathematical biology.
First, most mathematicians know less biology than
their tenth-grade daughters. That’s just the way it
is; biology was not a mathematically related disci-
pline when we were growing up. But, more impor-
tantly, mathematical biology really isn’t a “field” of
mathematics with a coherent community that can
testify meaningfully about young people. Mathe-
matical biology is fragmented because biology
itself is so diverse. A mathematician working on the
lung might want to talk to pulmonary physiologists
or geometric analysts who are experts on fractals,
but why would he or she want to talk to mathe-
matical biologists working on the kidney, the neu-
robiology of hearing, or the epidemiology of AIDS?
In each area of specialization there is a tremendous
amount of biology to learn, and if one doesn’t have
the background, it’s hard to judge the strength of
an individual’s contributions. This is true even for
us, the mathematical biologists, the “experts” you
expect to consult. So hiring in mathematical biol-
ogy necessarily involves intuition and high risks as
well as high potential payoffs for the department
and the college.
The first step. The best way for departments to
overcome these difficulties is to encourage senior
faculty to become involved in bringing biological
applications and student projects into the under-
graduate curriculum. This should be done by
working cooperatively with local biologists to
create examples and projects related to their
own specialties. All mathematicians can do this,
and they do not have to give up their own research
agendas or become mathematical biologists; it
only requires effort.
This strategy for engagement with biology has
great benefits both for departments and individu-
als. The faculty as a whole will become educated in
biology and thus better able to judge job candidates
in mathematical biology, and the undergraduate
curriculum will be more attractive. More importantly,
departments and individuals will be participating
intellectually in the biological revolution, the great-
est scientific revolution of our times, perhaps of all
times. The task is to understand how life, in all its
diversity and detail, works. This includes how we
act, think, and feel, and how we influence and are
influenced by other forms of life. We mathemati-
cians have the technical and intellectual tools to
make enormous contributions. So, surely, this is our
responsibility and our opportunity.

342 NOTICES OF THE AMS VOLUME 51, NUMBER 3