In partial fulfilment of the requirements of the course

Mathematical Modeling MATH F420

Modelling Selection Pressure in Evolution:

Selection Pressure exerted by Malaria on Sickle cell Gene

Prepared for and under the supervision of

Prof Dr. Anupama Sharma

(Dept of Mathematics, BITS Pilani K.K. Birla Goa Campus)

Prepared by:

Team Father Enigma

Abinidhi G 2020B4A81910G
Anuj Pradhan 2020A1PS2048G
Pranav Ghag 2020B4A32060G
Pranav Kocheta 2020B4A41859G
Sejal Sarada 2020B4A71849G
Shreya Agrawal 2020B4AA0785G
Stuti Nair 2020B4A40682G
 Abstract
The main goal of our project is to test the hypothesis that given lower malaria
death rates for heterozygous individuals, higher death rates due to malaria exert
selective pressure to extend the prevalence of the sickle cell gene for the country
Nigeria. Furthermore, we aim to find the numerical range of non carrier
individuals’ death caused due to malaria, i.e, the threshold value above which the
deadly trait of sickle cell starts to co-exist in the population, which we want to
prevent. This study derives a system of nonlinear differential equations with the
help of a compartmental model. We vary the parameter of death rate due to malaria
to find the coexistence between the different populations in our model for a span of
1000 years. Our results show that by keeping other parameters the same, raising
the adult,teen or child malaria death rate leads to a raised equilibrium proportions
of carrier populations showing that malaria exerts a selection pressure on the
human population to increase the prevalence of sickle cell.

2
 INDEX
Abstract 2
1. Introduction 4
2. The Model 6
2.1 Assumptions 6
2.2 Variables and Parameters 7
2.2.1 Variables 7
2.2.2 Parameters 7
2.3. Constructing the model 10
2.4. Formulating Equations 11
3. Analysis of the model 13
3.1. Jacobian matrix 13
3.2. Stability Analysis 14
3.3 Sensitivity analysis 16
3.3.1 Current Scenario 16
3.3.2. Malaria does not exist in the population 16
3.3.3. Malaria 0.5 times as deadly 17
3.3.4. Malaria 1.8 times as deadly 17
3.3.5. Malaria twice as deadly 18
3.3.6. Malaria is six times as deadly 18
4. Conclusion 20
5. Limitations 21
References 22
Appendix 23

3
 1. Introduction
Malaria is a mosquito borne disease that affects human beings which may result in
mild to acute illness depending upon the species that a person is infected by. Out of
the four species of malaria, Plasmodium Falciparum is considered to be one of the
most fatal among humans, particularly among children between the age of 0-5
years.Nigeria accounts for about 31.9% of the global malaria deaths; this is
approximately 200,000 deaths in 2021. Over 60 million people are infected yearly
and an estimated US$1.1 billion is lost yearly due to malaria related absenteeism
and productivity losses. (Guardian, 2022)
Sickle cell is a genetic disease caused by a recessive gene that deforms red blood
cells, resulting in severe anaemia among homozygous individuals and mild
symptoms among heterozygous carriers of the gene. Homozygous individuals are
people who have both the allele for sickle cell whereas heterozygous individuals
are the people with one allele for sickle cell and the other allele for healthy cell.
Homozygous non-carriers are individuals with no allele for sickle cell.(Mayoclinic)
Co-incidentally, reports suggest that the malaria parasite which during one part of
its life cycle infects red blood cells, increases the tendency of infected cells to
sickle, causing a very high mortality rate among homozygous individuals (Makani
et al. 2010).
However, several studies have suggested that the individuals who are carriers of
sickle cell disease, seem to offer resistance against malaria.(Luzzatto L. 2012). The
relative prevalence of sickle cell disease in countries suffering from malaria
suggests that malaria has exerted a selection pressure on the human population to
increase the prevalence of sickle cell.

4
 2. The Model
In our project, we have designed a compartmental model which is a simulation of
the real-world situation where carriers, non-carriers and people afflicted with sickle
cell disease across different age groups in Nigeria (i.e children, teens and adults)
respond to varying levels of prevalence of malaria.

2.1 Assumptions
The following are the assumptions we made while formulating our model:
1. Population is homogeneously mixed.
2. Malaria and sickle-cell anaemia both may be fatal.
3. Only the adult population, except afflicted ones, reproduce.
4. Carrying capacity is assumed to include all natural death rates for children
other than those specific to sickle cell/malaria. Natural death for teenagers
has been omitted.
5. Death rate due to sickle cell disease is equal for all age groups.
6. The population starts with adults, each one having equal probability of being
carrier, afflicted or non-carrier.
7. Over a span of 1000 years there is no development in Nigeria such that they
can treat sickle cell disease or malaria.Furthermore, we’ve also assumed that
malaria and sickle cell disease do not evolve over time.

5
2.2 Variables and Parameters

2.2.1 Variables
Our model incorporates three sickle cell genotypes: homozygous for sickle
cell gene (afflicted), homozygous for no sickle cell gene (non-carrier) and
heterozygous (non-carrier). We have further classified our model based on
different age groups: children (0-5 years), teenagers (5-15 years) and adults
(>15years). This classification has been done because the existing data
shows that malaria results in higher mortality and morbidity in children aged
0-5 years compared to teens and adults.(WHO, 2021) We get a model
consisting of nine different populations hence incorporating 9 variables.

Y: Adult non-carriers V:Child non-carriers F:Teen non-carriers

Z: Adult carriers W:Child carriers E:Teen carriers
X: Adult afflicted U:Child afflicted D:Teen afflicted

2.2.2 Parameters

The birth rate is found to be 3.64 percent per year (UN, 2019)

The natural death rate is computed by subtracting the deaths of adults due to
malaria and sickle cell disease from the total number of deaths during that
particular year, dividing it by total population and then multiplying it by 100.
Natural death rate = 1.14 - (0.75 + 0.044) = 0.346

Death rate due to malaria in adults, children and teens is computed by

obtaining the total number of deaths for that age group (Our World in Data,
2019) and dividing it by the total population during that time period.

85991
Death rate, adult, malaria = 200,963,599
* 100 = 0. 044

6
Death rate, teen, malaria = = 0.0047

Death rate, child, malaria = = 0.047

Death rate due to sickle cell disease in Nigeria was estimated to be in the
range of 50% to 80% (Muoghalu, 2018). We have chosen 75% i.e. 0.75 as the
most appropriate parameter value.

Maturation rate to teen, i.e from child to teen is 20% per year based on 5
years of childhood. Out of the five groups i.e. one,two,three, four and five
1
year olds only five year olds mature to teens so the maturation rate is 5
=0.2.

Similarly for maturation rate to adults out of the ten age groups (5-15 year
1
olds) only 15 year olds mature to adults, giving a maturation rate of 10
=0.1.

7
Symbol Parameter Value

𝜶 Birth rate 3.64

Natural Death rate* 0.346

𝛄N

Death rate, adult, malaria 0.0427

𝛄A,mal

Death rate, teen, malaria 0.0047

𝛄T,mal

Death rate, children, malaria 0.047

𝛄C,mal

Death rate due to sickle cell 0.75

𝛄Scd

Maturation rate to teen 0.2

mT

Maturation rate to adult 0.1

mA

Table 1: Parameter values

8
2.3. Constructing the model

Fig 1: This is a compartmental model incorporating 9 different populations based on age group and sickle
cell genotypes.

As mentioned in the paper above, the carrier population is observed to show

resistance against malaria disease. Hence, none of children, teen or adult in this
section of the population can die of malaria disease. The afflicted population is
homozygous to sickle cell gene therefore they can die of sickle cell disease. The
non-carriers being homozygous to no sickle cell gene, offers no resistance against
malaria disease hence, this section of the population is expected to die of malaria
disease. All the adults can also die naturally.

9
2.4. Formulating Equations

Where Bw Bu Bv are the birth terms shown below:

10
Fig 2: This schematic diagram represents the probability of reproduction of children of a particular sickle
cell genotype. It has been assumed that the afflicted adults cannot breed.

To obtain the birth terms, the general population growth rate, α, is multiplied
by the contribution from each genotype. For example, using a punnett square
it is known that a carrier children may result from a union of two carrier
adults or by a union of one carrier and one non-carrier adult.The probability
of having a carrier child is given by

This probability is together multiplied by the growth rate, the total number of
adult population i.e. (X+Y+Z) to get the proportions and a term that reflects
the carrying capacity of one (1-c) where c =(X+Y+Z+U+V+W+E+D+F).
This carrying capacity includes all the natural death rates for children and
teens other than those specific to malaria or sickle cell disease. Finally giving
us

11
3. Analysis of the model

3.1. Jacobian matrix

Setting the above nine equations to zero, the fixed points were found out .
The 9 variables in 9 equations are solved using a a python script. This led to
the procurement of one fixed point. The trivial fixed point (0,0,0,0,0,0,0,0,0)
was found out intuitively

The fixed point:

Using the above 9 equations we constructed a Jacobian matrix. After some

algebraic simplification involving row operations, we obtained the following
matrix.

Where A,B,C,D,E,F,G,H are:

12
3.2. Stability Analysis
In order to find the eigenvalues of the matrix, which would determine the
stability of the fixed point, the values of the parameters and the fixed point
were substituted. The complexity of the 9x9 matrix made it inefficient to
solve by hand and we thus resorted to a Octave script. On inputting the
parameter values, the following matrix is obtained.

From the above matrix, the following eigenvalues are obtained through the
same script.

13
 All the real part of the eigenvalues except 0.3672+0i are negative
suggesting that the fixed point is saddle.

When the initial condition is set to the fixed point, straight lines are acquired. This
confirms that the values obtained are that of a fixed point.
As they are numerically calculated there exists a slight blip at the start and a
perfectly straight line is not obtained.

14
3.3 Sensitivity analysis

3.3.1 Current Scenario

Fig 3: If the current parameters are substituted, our model predicts that SCD will be eradicated
from the population in 400 years.

3.3.2. Malaria does not exist in the population

Fig 4: If Malaria were to be eradicted from the population, our model predicts that SCD will be
eradicated from the population in 200 years.

15
 3.3.3. Malaria 0.5 times as deadly

Fig 5: If the death rate of malaria were to be half its current value, our model predicts that SCD
will be eradicated from the population in 300 years.

3.3.4. Malaria 1.8 times as deadly

Fig 6: If the death rate of malaria were to be 1.8 times its current value, our model predicts that
SCD will be eradicated from the population in 1000 years.

16
 3.3.5. Malaria twice as deadly

Fig 7: If the death rate of malaria were to be twice its current value, our model predicts that SCD
will remain in the population for over a 1000 years.

3.3.6. Malaria is six times as deadly

Fig 8: If the death rate of malaria were to be six times its current value, our model predicts that
majority of the population will be a carrier for SCD.

17
Using these graphs, we attempt to fulfill the two objectives with which we set out
to make our model. To do this, we varied the 𝛄C,mal value. We chose to do this
because this is the death rate due to malaria in children which is the highest of all 3
categories. This parameter affects the malaria-infected population size the most.

Our first objective was to test the hypothesis that given lower malaria death rates
for heterozygous individuals, higher death rates due to malaria exert selective
pressure to extend the prevalence of the sickle cell gene.
As we can see from figure 3, in the current scenario for the given value of 𝛄C,mal,the
non-carrier population is what thrives at the end of 1000 years. The carrier
population becomes extinct by the 400 year mark.
When we reduce the 𝛄C,mal value to zero, we can see from figure 4 that again
non-carrier population are thriving at the end of 1000 years. The carrier population
goes extinct much faster, just before the 200 year mark.
Increasing from 0 to 0.5 times the current 𝛄C,mal value, We can see that the same
things happens with no-carriers, and the carriers take a little over 200 years to go
extinct.
Taking 𝛄C,mal as twice its current value, we see a drastic change in trends. Along
with the non-carrier population thriving, we see carrier population also staying for
a significantly longer period. We see that they never reach zero in the span of 1000
years.
When the 𝛄C,mal value is made to be 6 times its current value, we see a reversal in
trends. For this we now see the carrier population thriving above the non-carrier
population. They do not go extinct over the given span of a 1000 years, and don’t
seem to zero anywhere in the timeline.
Hence, from the above analysis of the graphs using sensitivity analysis, we can see
that as the 𝛄C,mal value increases, the carrier for sickle-cell population takes a longer
time to become extinct. Hence we can see that when malaria death rate becomes
higher, it exerts a selection pressure on the population which favours the carrier
population, who thrive and the non-carrier population goes down in size.

18
We use the same graphs for our second objective as well. On varying 𝛄C,mal by a
factor of 1.8 it takes a thousand years for SCD to be eradicated. However, changing
it by a factor of 2 makes it so that it stays in the population beyond it. After
changing the value in this interval several times we found out that once the factor
crosses a threshold value of 1.9, SCD will not be eradicated. Thus this fulfills our
second objective.

19
4. Conclusion
1. The model supports the hypothesis that malaria exerts selective pressure to
favour the sickle cell gene and extends the prevalence of the genotype in
form of heterozygous individuals among the population.

2. When other factors are kept constant, raising the adult,teen or child malaria
death rate leads to a raised equilibrium proportions of carrier populations.

3. The threshold value obtained is 1.9. This means that if the death rate due to
malaria becomes 1.9 times it’s current value, sickle cell disease will remain
in the population for a 1000 years. However if malaria were to be eradicated
from the population, sickle cell disease would die out in 200 years.

20
5. Limitations
The model doesn’t consider sex as a factor. The birth terms also consider the case
when the offspring results from a union of two adults of the same sex, which
obviously doesn’t happen in the real world. An approach to solving this could be to
consider the sex ratio as a parameter or divide the population into further
compartments.

1. The is a complicated model as it gives a 9x9 Jacobian matrix which is

infeasible to solve by hand. Thus getting a closed form solution in terms of
the parameters is difficult. Thus we have substituted values for the
parameters and the solution was found out using some code.

2. The model also fails to consider the possibility of both the malaria parasite
and the sickle cell trait evolving over time, which might not be true
especially over a long period of time such as 1000 years which has been
considered in the model.

3. Due to the complexity of the model, it cannot be asserted that the fixed
points obtained are the only fixed points which exist for the system.

21
References

1) Chelsea Liddell, N. O.-B. (2014). A Mathematical Model of Sickle Cell

Genome Frequency in response to selective pressure from malaria. Society for
Mathematical Biology.

2) United Nations(World Population Prospects, 2019) World Population prospects,

https://population.un.org/wpp/

3)Our World in Data. (2019). Our World in Data. Retrieved from Our World in
Data: https://ourworldindata.org/malaria

4) Muoghalu, C. O. (2018). Sickle Cell Disease Child Mortality - A Silent

Epidemic in Nigeria: Issues in Political economy. Open Access Blood Research &
Transfusion Journal.

5)Macrotrends. (2019). Nigeria Birth Rate 1950-2022. Retrieved from

Macrotrends: https://www.macrotrends.net/countries/NGA/nigeria/birth-rate

6) Oncofertility Consortium. (2014). Sickle Cell Anemia. Retrieved from

Oncofertility Consortium:
https://oncofertility.msu.edu/resources/sickle-cell-anemia#

7) Guardian. (2022). The Guardian News. Retrieved from The Guardian:

https://guardian.ng/news/world-malaria-day-nigeria-records-200000-deaths-loses-n
646bn-yearly-to-malaria

8) Mayo Foundation for Medical Education and Research (MFMER):

https://www.mayoclinic.org/diseases-conditions/sickle-cell-anemia/symptoms-caus
es/syc-20355876#

22
9) Luzzatto L. (2012). Sickle cell anaemia and malaria. Mediterranean journal of
hematology and infectious diseases, 4(1), e2012065.
https://doi.org/10.4084/MJHID.2012.065

10) World Health Organization (2021) Malaria fact sheet.

https://www.who.int/news-room/fact-sheets/detail/malaria

23
Appendix
Following are the scripts that we used.

1) For plotting graphs (Python)

import numpy as np
import matplotlib.pyplot as plt
from scipy.integrate import odeint

#Parameters

alpha = 3.64
gamma_scd = 0.75
gamma_A_mal = 0.042
gamma_T_mal = 0.0047
gamma_C_mal = 0.047
gamma_natural = 0.346
mt = 0.2
ma = 0.1

initals = [0, 0, 0, 1/3, 1/3, 1/3, 0, 0,0]

# u v w x y z d e f

#function
def Disease(x,t):
u = x[0] #Child Afflicted
v = x[1] #Child Non Carrier
w = x[2] #Child Carrier
X = x[3] #Adult Afflicted
y = x[4] #Adult Non Carrier
z = x[5] #Adult Carrier
d = x[6] #Teen afflicted
e = x[7] #Teen Carrier
f = x[8] #Teen Non Carrier

dx = np.zeros(9)

dx[0] = 0.25*alpha*z*z*(1-(u+v+w+X+y+z+d+e+f))/((X+y+z)) - mt*u - gamma_scd*u

dx[1] = alpha*(1-(u+v+w+X+y+z+d+e+f))*(y*y/(X+y+z) + 0.5*y*z/(X+y+z) +
0.25*z*z/(X+y+z) ) - mt*v - gamma_C_mal*v
dx[2] = alpha*(0.5*z*z/(X+y+z) + 0.5*y*z/(X+y+z))*(1-(u+v+w+X+y+z+d+e+f)) - mt*w

24
 dx[3] = ma*d - gamma_scd*X - gamma_natural*X
dx[4] = ma*f - gamma_A_mal*y - gamma_natural*y
dx[5] = ma*e - gamma_natural*z

dx[6] = mt*u - ma*d - gamma_scd*d

dx[7] = mt*w - ma*e
dx[8] = mt*v - ma*f - gamma_T_mal*f
return dx

#solving odes
t = np.array(np.linspace(0,1000,50000))
x = np.array(odeint(Disease,initals,t))

u = x[:,0]
v = x[:,1]
w = x[:,2]
X = x[:,3]
y = x[:,4]
z = x[:,5]
d = x[:,6]
e = x[:,7]
f = x[:,8]
N = X+y+z+u+v+w+d+e+f

#plotting
plt.figure(figsize=(8,5))

plt.plot(t,u,label = 'Child Afflicted')

plt.plot(t,v,label = 'Child Non Carrier')
plt.plot(t,w,label = 'Child Carrier')
plt.plot(t,X,label = 'Adult Afflicted')
plt.plot(t,y,label = 'Adult Non Carrier')
plt.plot(t,z,label = 'Adult Carrier')
plt.plot(t,d,label = 'Teen afflicted')
plt.plot(t,e,label = 'Teen Carrier')
plt.plot(t,f,label = 'Teen Non Carrier')
plt.plot(t,N,label = 'Total')
plt.legend(loc="best")

plt.xlabel("Years")
plt.ylabel("Fraction")
plt.show()

25
2) For finding the fixed point (Python)

from scipy.optimize import fsolve

def equations(vars):
alpha = 3.64
gamma_scd = 0.75
gamma_A_mal = 0.042
gamma_T_mal = 0.0047
gamma_C_mal = 0.047
gamma_natural = 0.346
ma = 0.1
mt = 0.2

u,v,w,X, y,z,d,e,f = vars

eq1 = 0.25*alpha*z*z*(1-(u+v+w+X+y+z+d+e+f))/((X+y+z)) - mt*u - gamma_scd*u

eq2 = alpha*(1-(u+v+w+X+y+z+d+e+f))*(y*y/(X+y+z) + 0.5*y*z/(X+y+z) +
0.25*z*z/(X+y+z) ) - mt*v - gamma_C_mal*v
eq3 = alpha*(0.5*z*z/(X+y+z) + 0.5*y*z/(X+y+z))*(1-(u+v+w+X+y+z+d+e+f)) - mt*w

eq4 = ma*d - gamma_scd*X - gamma_natural*X

eq5 = ma*f - gamma_A_mal*y - gamma_natural*y
eq6 = ma*e - gamma_natural*z

eq7 = mt*u - ma*d - gamma_scd*d

eq8 = mt*w - ma*e
eq9 = mt*v - ma*f - gamma_T_mal*f
return [eq1, eq2,eq3,eq4,eq5,eq6,eq7,eq8,eq9]

u,v,w,x,y,z,d,e,f = fsolve(equations,(1,1,1,1,1,1,1,1,1))

print(f"""u = {u}
v = {v}
w = {w}
x = {x}
y = {y}
z = {z}
d = {d}
e = {e}
f = {f}""")

26
3) To find eigen values (Octave)

#Parameters
alpha = 3.64
gamma_scd = 0.5
gamma_A_mal = 0.042
gamma_T_mal = 0.0047
gamma_C_mal = 0.047
gamma_natural = 0.346
ma = 0.1
mt = 0.2

#Fixed Point
u=0
v = 0.25338977070702223
w=0
x=0
y = 0.1247500323491873
z=0
d=0
e=0
f = 0.48403012551484664
n = 1-(x+y+z+u+v+w+d+e+f)

J11 = -mt
J12 = 0
J13 = 0
J14 = mt+ma
J15 = 0
J16 = 0
J17 = -ma
J18 = 0
J19 = 0
J21 = 0
J22 = -mt-gamma_scd
J23 = 0
J24 = ma
J25 = mt
J26 = 0
J27 = -ma
J28 = 0
J29 = 0
J31 = 0

27
J32 = 0
J33 = -mt-gamma_C_mal
J34 = ma
J35 = mt
J36 = 0
J37 = -ma
J38 = 0
J39 = 0
J41 = -(alpha/((x+y+z)*(x+y+z))) * (0.5*y*z+0.5*z*z)
J42 = -(alpha/((x+y+z)*(x+y+z))) * 0.25*z*z
J43 = -(alpha/((x+y+z)*(x+y+z)))*(y*y+0.5*y*z+0.25*z*z)
J44 = -ma
J45 = 0
J46 = 0
J47 = ma
J48 = 0
J49 = 0
J51 = 0
J52 = 0
J53 = 0
J54 = ma
J55 = -ma-gamma_scd
J56 = 0
J57 = -ma
J58 = ma
J59 = 0
J61 = 0
J62 = 0
J63 = 0
J64 = ma
J65 = 0
J66 = -ma-gamma_T_mal
J67 = -ma
J68 = 0
J69 = ma
J71 = (alpha*n/((x+y+z)*(x+y+z))) * (0.5*y+z) - (0.5*y*z + 0.5*z*z)
J72 = (alpha*n/((x+y+z)*(x+y+z)))*(0.5*z*(x+y+z)-0.25*z*z)
J73 =(alpha*n/((x+y+z)*(x+y+z)))*(0.5*(y+z)*(x+y+z)-(y*y+0.5*y*z+0.25*z*z))
J74 =ma
J75 = 0
J76 =0
J77 =-ma-gamma_T_mal
J78 =0
J79 =0

28
J81 =-(alpha*n/((x+y+z)*(x+y+z)))*(0.5*y*z+0.5*z*z)
J82 =(alpha*n/((x+y+z)*(x+y+z)))*(-0.25*z*z)
J83 =(alpha*n/((x+y+z)*(x+y+z)))*(y*y+0.5*y*z+0.25*z*z)
J84 =ma
J85 =0
J86 =0
J87 =-ma
J88 =-gamma_scd
J89 =0
J91 =(alpha*n/((x+y+z)*(x+y+z)))*((x+y+z)*0.5*z-(0.5*y*z+0.5*z*z))
J92 =(alpha*n/((x+y+z)*(x+y+z)))*(-0.25*z*z)
J93 =(alpha*n/((x+y+z)*(x+y+z)))*((x+y+z)*(2*y+0.5*z)-(y*y+0.5*y*z+0.25*z*z))
J94 =ma
J95 =0
J96 =0
J97 =-ma
J98 =0
J99 =-gamma_A_mal-gamma_natural

J = [J11 J12 J13 J14 J15 J16 J17 J18 J19;

J21 J22 J23 J24 J25 J26 J27 J28 J29;
J31 J32 J33 J34 J35 J36 J37 J38 J39;
J41 J42 J43 J44 J45 J46 J47 J48 J49;
J51 J52 J53 J54 J55 J56 J57 J58 J59;
J61 J62 J63 J64 J65 J66 J67 J68 J69;
J71 J72 J73 J74 J75 J76 J77 J78 J79;
J81 J82 J83 J84 J85 J86 J87 J88 J89;
J91 J92 J93 J94 J95 J96 J97 J98 J99]

lamda = eig(J)

29