Part A

Click to add Bookmark this page

Part A
8.0 points possible (graded)

PART A: PROBABILITIES
This part of the problem set involves some pencil-and-paper exercises. It will
help you practice and understand simple probability and statistics.
Let's say Alvin will catch the flu with probability of 1/10 during any given month.
Let's also assume that Alvin can catch the flu only once per month, and that if he
has caught the flu, the flu virus will die by the end of the month. What is the
probability of the following events?
Answer each question in reduced fraction form - eg 1/5 instead of 2/10.
1.

He catches the flu in September, October and November.

unanswered

2.

He catches the flu in September and then again in November, but not in
October.
unanswered

3.

He catches the flu exactly once in the three months from September
through November.
unanswered

4.

He catches the flu in two or more of the three months from September
through November.
unanswered
Part A

Save Save your answer ResetReset your answer

SubmitSubmit your answer
You have used 0 of 10 attempts

INTRODUCTION
In this problem set, using Python and Pylab, you will design and implement a
stochastic simulation of patient and virus population dynamics, and reach
conclusions about treatment regimens based on the simulation results.

GETTING STARTED
Download: ProblemSet3.zip, a skeleton file for Part B.

BACKGROUND: VIRUSES, DRUG TREATMENTS, AND

COMPUTATIONAL MODELS
Viruses such as HIV and H1N1 represent a significant challenge to modern
medicine. One of the reasons that they are so difficult to treat is their ability to
evolve.
As you may know from introductory biology classes, the traits of an organism
are determined by its genetic code. When organisms reproduce, their offspring
will inherit genetic information from their parent. This genetic information will
be modified, either because of mixing of the two parents' genetic information, or
through mutations in the genome replication process, thus introducing diversity
into a population.
Viruses are no exception. Two characteristics of viruses make them particularly
difficult to treat. The first is that their replication mechanism often lacks the
error checking mechanisms that are present in more complex organisms. This
speeds up the rate of mutation. Secondly, viruses replicate extremely quickly
(orders of magnitude faster than humans) -- thus, while we may be used to
thinking of evolution as a process which occurs over long time scales,
populations of viruses can undergo substantial evolutionary changes within a
single patient over the course of treatment.
These two characteristics allow a virus population to acquire genetic resistance
to therapy quickly. In this problem set, we will make use of simulations to
explore the effect of introducing drugs on the virus population and determine
how best to address these treatment challenges within a simplified model.
Computational modeling has played an important role in the study of viruses
such as HIV (for example, see this paper, by MIT graduate David Ho). In this
problem, we will implement a highly simplified stochastic model of virus
population dynamics. Many details have been swept under the rug (host cells
are not explicitly modeled and the size of the population is several orders of
magnitude less than the size of actual virus populations). Nevertheless, our
model exhibits biologically relevant characteristics and will give you a chance to
analyze and interpret interesting simulation data.

SPREAD OF A VIRUS IN A PERSON

In reality, diseases are caused by viruses and have to be treated with medicine,
so in the remainder of this problem set, we'll be looking at a detailed simulation
of the spread of a virus within a person. We've provided you with skeleton code
in ps3b.py.

Part B: Problem 1: Implementing a Simple Simulation (No Drug

Treatment)
15.0 points possible (graded)

We start with a trivial model of the virus population - the patient does not take
any drugs and the viruses do not acquire resistance to drugs. We simply model
the virus population inside a patient as if it were left untreated.

SIMPLEVIRUS CLASS
To implement this model, you will need to fill in the SimpleVirus class, which
maintains the state of a single virus particle. You will implement the methods
__init__, getMaxBirthProb, getClearProb,doesClear, and reproduce
according to the specifications. Use random.random() for generating random
numbers to ensure that your results are consistent with ours.
Hint: random seed

During debugging, you might want to use random.seed(0) so that your results
are reproducible.
The reproduce method in SimpleVirus should produce an offspring by
returning a new instance of SimpleVirus with probability:
self.maxBirthProb * (1 - popDensity). This method raises a
NoChildException if the virus particle does not reproduce. For a reminder on
raising execptions, review the Python docs.
self.maxBirthProb is the birth rate under optimal conditions (the virus
population is negligible relative to the available host cells so there is ample
nourishment available). popDensity is defined as the ratio of the current virus
population to the maximum virus population for a patient and should be
calculated in the update method of the Patient class.

PATIENT CLASS
You will also need to implement the Patient class, which maintains the state of
a virus population associated with a patient.
The update method in the Patient class is the inner loop of the simulation. It
modifies the state of the virus population for a single time step and returns the
total virus population at the end of the time step. At every time step of the
simulation, each virus particle has a fixed probability of being cleared
(eliminated from the patient's body). If the virus particle is not cleared, it is
considered for reproduction. The virus population should never exceed maxPop;
if you utilize the population density correctly, you shouldn't need to provide an
explicit check for this.

Unlike the clearance probability, which is constant, the probability of a virus

particle reproducing is a function of the virus population. With a larger virus
population, there are fewer resources in the patient's body to facilitate
reproduction, and the probability of reproduction will be lower. One way to think
of this limitation is to consider that virus particles need to make use of a
patient's cells to reproduce; they cannot reproduce on their own. As the virus
population increases, there will be fewer available host cells for viruses to utilize
for reproduction.
To summarize, update should first decide which virus particles are cleared and
which survive by making use of the doesClear method of each SimpleVirus
instance, then update the collection of SimpleVirus instances accordingly. With
the surviving SimpleVirus instances, update should then call the reproduce
method for each virus particle. Based on the population density of the surviving
SimpleVirus instances, reproduce should either return a new instance of
SimpleVirus representing the offspring of the virus particle, or raise a
NoChildException indicating that the virus particle does not reproduce during
the current time step. The update method should update the attributes of the
patient appropriately under either of these conditions. After iterating through all
the virus particles, the update method returns the number of virus particles in
the patient at the end of the time step.
Hint: mutating objects

Be very wary about mutating an object while iterating over its elements. It is best
to avoid this entirely (consider introducing additional "helper" variables). See the
6.00.2x Style Guide for more information.
Note that the mapping between time steps and actual time will vary depending
on the type of virus being considered, but for this problem set, think of a time
step as a simulated hour of time.
About the grader: When defining a Patient class member variable to store the
viruses list representing the virus population, please use the name
self.viruses in order for your code to be compatible with the grader and to
pass one of the tests.

Part B: Problem 2
Click to add Bookmark this page

Part B: Problem 2-1: Running and Analyzing a Simple

Simulation (No Drug Treatment)

10.0 points possible (graded)

You should start by understanding the population dynamics before introducing
any drug.
Fill in the function simulationWithoutDrug(numViruses, maxPop,
maxBirthProb, clearProb, numTrials) that instantiates a Patient,
simulates changes to the virus population for 300 time steps (i.e., 300 calls to
update), and plots the average size of the virus population as a function of time;
that is, the x-axis should correspond to the number of elapsed time steps, and
the y-axis should correspond to the average size of the virus population in the
patient. The population at time=0 is the population after the first call to update.
Run the simulation for numTrials trials, where numTrials in this case can be
up to 100 trials. Use pylab to produce a plot (with a single curve) that displays
the average result of running the simulation for many trials. Make sure you run
enough trials so that the resulting plot does not change much in terms of shape
and time steps taken for the average size of the virus population to become
stable. Don't forget to include axes labels, a legend for the curve, and a title
on your plot.
You should call simulationWithoutDrug with the following parameters:
numViruses = 100
maxPop (maximum sustainable virus population) = 1000
maxBirthProb (maximum reproduction probability for a virus particle) =
0.1
clearProb (maximum clearance probability for a virus particle) = 0.05
Thus, your simulation should be instantiatating one Patient with a list of 100
SimpleVirus instances. Each SimpleVirus instance in the viruses list should
be initialized with the proper values for maxBirthProb and clearProb.
Hint: graphing
Your graph should contain one line that represents the average of many
different trials. One way of computing the average involves holding all of your
data in one list, with one element for each of 300 time steps, and adding to each
data point during each trial. Then, at the end, each element of the list is divided
by the total number of trials, thus taking the average of each element of the list.
However, if this idea confuses you, feel free to ignore the hint and implement it
however makes the most sense to you.
Consult reference documentation on pylab as reference. Scroll down on the
page to find a list of all the plotting commands in pylab.
Hint: testing your simulation

Compared to the previous problem sets, testing your simulation code is more
challenging, because the behavior of the code is stochastic, and the expected
output is not exactly known. How do you know whether your plots are correct or
not? One way to test this is to run the simulation with extreme input values (i.e.
initialization parameters), and check that the output matches your intuition. For
example, if maxBirthProb is set to 0.99 instead of 0.1, then you would expect
that the virus population rapidly increases over a short period of time. Similarly,
if you run your simulation with clearProb = 0.99 and maxBirthProb = 0.1,
then you should see the virus population quickly decreasing within a small
number of steps. You can also try to vary the input values, and check whether
the output plots change as you expect. For example, if you run multiple
simuation runs, each time increasing maxBirthProb, the curves in the
successive plots should show an "upward" trend, since the virus will reproduce
faster with a higher maxBirthProb.
For further testing, we have provided the .pyc (compiled Python) files for the
completed Patient and SimpleVirus classes (and for Problem 5, the
ResistantVirus and TreatedPatient classes) that you can use to confirm
that your code is generating the correct results during simulation.
If you comment out your versions of these classes in ps3b.py, and add the
following import statements to the top of your file, you can run the simulation
using our pre-compiled implementation of these classes to make sure you are
obtaining the correct results. This is a good way to test if you've implemented
these classes correctly. Make sure to comment out the import statement and
uncomment your implementations before moving to Problem 4.
For Python 3.5: from ps3b_precompiled_35 import *

Part B: Problem 3
Click to add Bookmark this page

Part B: Problem 3: Implementing a Simulation With Drugs

10.0 points possible (graded)

In this problem, we consider the effects of both administering drugs to the

patient and the ability of virus particle offsprings to inherit or mutate genetic
traits that confer drug resistance. As the virus population reproduces, mutations
will occur in the virus offspring, adding genetic diversity to the virus population.
Some virus particles gain favorable mutations that confer resistance to drugs.

RESISTANTVIRUS CLASS
In order to model this effect, we introduce a subclass of SimpleVirus called
ResistantVirus. ResistantVirus maintains the state of a virus particle's
drug resistances, and accounts for the inheritance of drug resistance traits to
offspring. Implement the ResistantVirus class.

Part B: Problem 4
Click to add Bookmark this page

Part B: Problem 4: TreatedPatient Class

10.0 points possible (graded)

We also need a representation for a patient that accounts for the use of drug
treatments and manages a collection of ResistantVirus instances. For this, we
introduce the TreatedPatient class, which is a subclass of Patient.
TreatedPatient must make use of the new methods in ResistantVirus and
maintain the list of drugs that are administered to the patient.
Drugs are given to the patient using the TreatedPatient class's
addPrescription() method. What happens when a drug is introduced? The
drugs we consider do not directly kill virus particles lacking resistance to
the drug, but prevent those virus particles from reproducing (much like actual
drugs used to treat HIV). Virus particles with resistance to the drug continue to
reproduce normally. Implement the TreatedPatient class.
Hint: reproduce function child resistances
If you are really unsure about how to think about what each child resistances
should be changed to, here is a different approach. If the probability mutProb is
successful, the child resistance switches. Otherwise, the child resistance stays
the same as the parent resistance.

Part B: Problem 5
Click to add Bookmark this page

Part B: Problem 5: Running and Analyzing a Simulation With a

Drug

10.0 points possible (graded)

In this problem, we will use the implementation you filled in for Problem 4 to run
a simulation. You will create a TreatedPatient instance with the following
parameters, then run the simulation:
viruses, a list of 100 ResistantVirus instances
maxPop, maximum sustainable virus population = 1000
Each ResistantVirus instance in the viruses list should be initialized with the
following parameters:
maxBirthProb, maximum reproduction probability for a virus particle =
0.1
clearProb, maximum clearance probability for a virus particle = 0.05
resistances, The virus's genetic resistance to drugs in the experiment =
{'guttagonol': False}
mutProb, probability of a mutation in a virus particle's offspring = 0.005
Run a simulation that consists of 150 time steps, followed by the addition of the
drug, guttagonol, followed by another 150 time steps. You should make use of
the function simulationWithDrug(numViruses, maxPop, maxBirthProb,
clearProb, resistances, mutProb, numTrials). As with problem 3,
perform up to 100 trials and make sure that your results are repeatable and
representative.
Create one plot that records both the average total virus population and the
average population of guttagonol-resistant virus particles over time.
A few good questions to consider as you look at your plots are: What trends do
you observe? Are the trends consistent with your intuition? Feel free to discuss
the answers to these questions in the forum, to fully cement your understanding
of this problem set, processing and interpreting data.
Again, as in Problem 2, you can use the provided .pyc file to check that your
implementation of the TreatedPatient and ResistantVirus classes work as
expected.