WICI Waterloo 2016
Infectious disease modelling over many scales Jane Heffernan York Research Chair Centre for Disease Modelling Modelling Infection and Immunity Mathematics & Statistics York University
Multiple Scales
Modelling Infection and Immunity Lab How does the immune system interact with a pathogen? How do you develop immunity? How is this related to transmission? How does some population immunity affect epidemic size?
2 MI
Multi-scale In-Host
Projects
Between Host
HIV
Pop’n Level
Experimental
HIV
HBV/HCV Measles
Measles
Measles Pertussis
Influenza
Influenza
Influenza
TB
TB
TB
Multiple Exposure
Multiple Exposure
Multiple Exposure
Influenza
Herpes Amyloidosis
Amyloidosis Social Distancing
Dynamical systems, stochastic models, computer simulations, bifurcations, stability
Background Mathematical Epidemiology Mathematical Immunology Basic Reproductive Ratio
Epidemiology S
E
I
R
S - susceptible dS = λ − d S S − βSI E – exposed dt dE I – infectious (symptomatic = βSI − d E E − αE dt for part of the time) dI = αE − d I I − γI R - recovered dt λ - birth rate dR = γI − d R R α, γ – disease progression dt β – infection rate Equilibria – uninfected, infected di - death rate
Epidemiology Basic reproductive ratio The number of secondary infections produced by an initial infective in a totally susceptible population
Threshold condition – transcritical bifurcation (usually)
Epidemiology Basic reproductive ratio The number of secondary infections produced by an initial infective in a totally susceptible population
dS = λ − d S S − βSI dt dE = βSI − d E E − αE dt dI = αE − d I I − γI dt dR = γI − d R R dt
What can be measured?
Extra Complexity V
S
Ev
E
Iv
Rv
IT
RT
I
R
Sw
Epidemiology Model with drugs S
E
I
Extensions dS = λ − d S S − (1 − ε ) βSI Metapopulations dt dE = (1 − ε ) βSI − d E E − αE Networks dt Age dI = αE − d I I − γI Space dt dR And so on… = γI − d R dt
R
R Other compartments Vaccinated Partially immune Asymptomatic infectious Cancer Chronic Infection And so on…
General Disease and Immune Stages All states should depend on immune system status birth
S
infection
E*
* death
C*
recovery progression progression
death
I*
death
V*
waning immunity
W*
vaccination death
recovery
R* death
waning immunity
infection vaccination
death
death
Immune System
Cells of Immune System
• Naive, activated, memory
• T – produced by thymus • CD4 – Helper T-cells – Activated by APCs – Activate the immune system
• CD8
– Killer T-cells
• B - produced in bone marrow • Mature in spleen • Plasma cells, antibodies
Pathogenesis Establishment of infection
Initiate adapative immune response
Adapative immune response
Immunological memory
Level of pathogen in plasma
Immune memory
Level of immunity
Chronic Infection
Acute Infection
Pathogen enters body
Pathogen enters plasma
Infectiousness begins
Symptoms appear
Infectiousness: little to none
Pathogen: clearance or persistence
Basic Model HIV
Protease Fusion
Reverse Transcriptase Viral RNA Viral RNA Transcribed to DNA
RNA + Viral Proteins Released
Viral DNA Incorporated Into Host Genome
New Proteins from Viral DNA
CD4 receptor
Budding of New Virion Protease Enables Capsid Assembly
CD4 T-cell
Basic Model dx = λ − d x x − βxv dt dy = βxv − d y y dt dv = ky − d v v dt
x - uninfected cells y - infected cells v - free virus Nowak, M.A. and R.M. May, Virus Dynamics
λ, k - production rate β - efficacy of infection dx, dy, dv - death rates/ clearance time
Basic Model- with virus loss dx = λ − d x x − βxv dt dy = βxv − d y y dt dv = ky − d v v − βxv dt
x - uninfected cells y - infected cells v - free virus
λ, k - production rate β - efficacy of infection dx, dy, dv - death rates/ clearance time
Basic Model- with virus loss dx = λ − d x x − βxv dt dy = βxv − d y y dt dv = ky − d v v − βxv dt
dx = λ − d x x − β xv dt dE ( X ) = λ − d x E ( X ) − β E ( XV ) dt E ( XV ) = E ( X ) E (V ) + COV ( X , V )
Moment closure
In-host Model Basic reproductive ratio The number of secondary infections produced by an initial infective in a totally susceptible population
βx0 k R0 = d v + βx0 d y
In-host Model Basic reproductive ratio The number of secondary infections produced by an initial infective in a totally susceptible population
βx0 k R0 = d v + βx0 d y
Basic Model- with drugs dx = λ − d x x − (1 − ε rt ) βxv dt dy = (1 − ε rt ) βxv − d y y dt dv = (1 − Qε p )ky − d v v − βxv dt dw = ε p Qky − d v w dt
x - uninfected cells y - infected cells v, w - free virus
λ, k - production rate β - efficacy of infection dx, dy, dv - death rates/ clearance time
Extensions to Basic Model Drug therapy Inhibit infection of a cell by virus Inhibit production of new virions Pharmacokinetics – doses of drugs, adherence
Incorporate (some of) the immune system?
Activation, development of memory CD8 cells, killer T-cells Antibodies and B-cells Innate immune system – cytokines
Spatial effects – location of infection And many more
CD8 T-cell immunity death ‘birth’
death Infection
x
y
budding
PBMCs – x, y
v+w proliferation
CD8’s – z conversion
zn
za activation
death
zm activation death
death
Initiate adapative immune response
CD8 T-cell immunity Level of pathogen in plasma
Establishmen t of infection
Level of death immunity Infection ‘birth’
x
Immunological memory
Adapative immune response
death
y
budding Chronic Infection
PBMCs – x, y
Pathogen enters body
CD8’s – z
Immune memory
v+w
Acute Infection
Pathogen enters plasma
Infectiousness begins
proliferation
Symptoms appear
Infectiousness: little to none
Pathogen: clearance or persistence
conversion
zn
za activation
death
zm activation death
death
CD4 T-cell immunity death ‘birth’
death conversion
activation
xn
death
xa
activation
xm
proliferation
v+w
proliferation
Infection by infectious virus v
budding
yn
ya activation
death
conversion
ym
activation death
death
CD4 T-cell immunity death
death
death conversion
activation
Longlived!!
‘birth’
xn
xa
activation
xm
proliferation
proliferation
v+w
Infection by infectious virus v
budding
yn
ya
activation
death
conversion
ym
activation death
death
HIV and Latency death ‘birth’
death conversion
activation
xn
death
xa
activation
xm
proliferation
v+w
proliferation
Infection by infectious virus v
budding
yn
ya activation
death
conversion
ym
activation death
death
Modelling Pathogen Dynamics and Immune System Memory In-Host Cao P, Yan AWC, et al. (2015) PLoS Comp. Biol., 11(8) e1004334. Laurie KL, Guarnaccia TA, et al. (2015) JID, published online jiv260. Du Y, Wu J, Heffernan JM (2014). Math. Pop. Stud., in press. Frascoli F, Wang Y, et al. (2014). CAMQ, in press. Wang Y, Brauer F, et al. (2014) Jour Math Analysis Appl, 414(2), 514-531 Wang Y, Zhou Y, et al. (2013) Jour Math Biol., 67(4), 901-934. Qesmi R, ElSaadany S, et al. (2011), SIAM J. Appl. Math. 71, 1509-1530. Qesmi R, Wu J, et al. (2010) Math Biosci. 4(2), 118-25. Heffernan JM, Keeling MJ (2008). Theor. Popul. Biol., 73, 134-147. Heffernan JM, Wahl L.M (2006). Jour. Theor. Biol., 243(2), 191-204.
HIV and Latency death ‘birth’
death conversion
activation
xn
death
xa
activation
xm
proliferation
v+w
proliferation
Infection by infectious virus v
budding
yn
ya activation
death
conversion
ym
activation death
death
Viral Blips  ODEs, include long lived memory cells
Frascoli F, Wang Y, et al. (2014). CAMQ, in press.
Viral Blips ODEs, include long lived memory cells
Tat ‘on-off’ switch On-going MI2 work Bifurcations: Backward (1 or 2), Hopf (1 or 2), depend on s anf f(x,y,v)Tat ‘on-off’ switch
Frascoli F, Wang Y, et al. (2014). CAMQ, in press.
Acute - Influenza Innate and B-cells – reinfection intervals Ferrets Viral shedding Nasal wash – 24hrs
Viral hierarchies Prevention of 2nd inf Co-inf Shortened 2nd inf Delayed 2nd inf No effect
C
Cao P, Yan AWC, et al. (2015) PLoS Comp. Biol., accepted. Laurie KL, Guarnaccia TA, et al. (2015) JID, published online jiv260.
10
Block/prevention A(H1N1)pdm09
log10 copy number / 100 µl nasal wash
10
Co-infection
10
A(H3N2) A(H1N1)pdm09
B
8
8
8
6
6
6
4
4
4
10
Delay A(H1N1)pdm09
10
B 8
6
6
4
4
Control
No effect A(H1N1)pdm09
B
8
10
10
limit of detection of infectious virus
Control
10 A(H1N1)pdm09
B 8
8
6
6
6
0 2 4 6 8 10 12 14 16 18 20 22
4
0 2 4 6 8 10 12 14 16 18 20 22
experimental day
Control A(H3N2)
8
4
Shortened
A(H1N1)pdm09
B
4
0 2 4 6 8 10 12 14 16 18 20 22
Acute - Influenza Innate and B-cells – viral hierarchies
• • • •
productive co-infection (grey) early synchronised decrease (red) Desynchronised phase (green) removal of the challenge virus (blue)
Cao P, Yan AWC, et al. (2015) PLoS Comp. Biol., accepted. Laurie KL, Guarnaccia TA, et al. (2015) JID, published online jiv260.
Acute – Measles death ‘birth’
death Infection
x
y
Heffernan JM, Keeling MJ (2008). Theor. Popul. Biol., 73, 134-147.
budding
PBMCs – x, y
v+w proliferation
CD8’s – z conversion
zn
za activation
death
zm activation death
death
Acute - Measles CD8 T-cell – natural infection and vaccination Heffernan JM, Keeling MJ (2008). Theor. Popul. Biol., 73, 134-147.
Acute - Measles CD8 T-cell – natural infection and vaccination Subclinical infection
Translation to epidemiology
Heffernan JM, Keeling MJ (2008). Theor. Popul. Biol., 73, 134-147.
Natural Inf + Booster Infs Subclinical Infection
37
Vaccine + Booster Infections Subclinical Infection
38
Modelling Immunity in a Population Qesmi R, Heffernan JM, Wu J (2015) Jour. Math. Biol. 70 (1-2), 343-366. Dafilis M, Frascoli F, et al. (2014) Theor Biol Med Model 11:43. Dafilis M, Frascoli F, et al. (2014) Jour Theor Biol 361, 124-132. Lou Y, Steben M, et al. (2014) Discrete and Continuous Dynamical Systems B 19(2), 447-466. Duvvuri VRSK, Heffernan JM, et al. (2012) BMC Inf Dis, 12:329 Lou Y, Qesmi R, et al. (2012) PLoS One, e46027. Ghosh S, Heffernan JM (2010). PLoS One, e14307. Heffernan JM, Keeling MJ (2009). Proc Roy Soc B, 276(1664), 2071-2080.
Waning Immunity - Measles ω p −k , ..., ωi
∑ β jY j σ → γ ω , ..., ω αi j p p p −k +1 Si → Ei → Yi → R p → S p −k Parameterize epidemiological model using in-host output
β i , α i , γ i , ωi , σ
Other Parameters Host natural death rate Host immunity vaccination distribution Heffernan JM, Keeling MJ (2009). Proc Roy Soc B, 276(1664), 2071-2080.
40
Waning Immunity - Measles No vaccine Distribution of Immunity
Heffernan JM, Keeling MJ (2009). Proc Roy Soc B, 276(1664), 2071-2080.
Symptomatic Infections
Function of age Variation from standard SEIR is slight and primarily occurs as a mild infection in older individuals
Vaccination + Waning Distribution of immunity 30 and 80 years
Avg Prevalence of Infection Waning immunity can severely limit effects of vaccination – L vs. NL
not linear linear
42 Heffernan JM, Keeling MJ (2009). Proc Roy Soc B, 276(1664), 2071-2080.
Vaccination High levels of vaccination (>70%) and moderate levels of waning immunity (>30 years) lead to large scale epidemic cycles 92 % vacc 30,40,50,60 years waning immunity
43 Heffernan JM, Keeling MJ (2009). Proc Roy Soc B, 276(1664), 2071-2080.
Hopf, Medium, Large
Waning Immunity
What if immunity does not decay to zero? Consider waning to min of 6 mem T-cells /μL Can do similar analysis Not very different Level of infectious cases is higher, but more are asymptomatic Magnitude and period is smaller for outbreaks Suboptimal boosting 45
SIMPLE SIRWS Model Waning immunity (1/κ=10 yr) and immune boosting (v) Dafilis M, Frascoli F, et al. (2014) Theor Biol Med Model 11:43. Dafilis M, Frascoli F, et al. (2014) Jour Theor Biol 361, 124-132. • Period diagrams (η,ν)-plane • 1/ξ=50, 55, 60, 65, 70, 75, 80, 85, 90, 95, 100 years • saddle node lines and period-doubling cascades of different orbital periods overlap • system sensitive to small perturbations in parameters and prone 46 to multistable behaviour
SIMPLE SIRWS Model 1/1 1/2 1/3
2/4 2/5 3/6
4/6 4/10 6/12
Dafilis M, Frascoli F, et al. (2014) Theor Biol Med Model 11:43. Dafilis M, Frascoli F, et al. (2014) Jour Theor Biol 361, 124-132.
SIMPLE SIRWS Model  Basins of attraction, vary duration of immunity
Period of Oscillations
Infecteds
2D slices through the full 4D initial-condition space. S(0)+I(0)+R(0)+W(0) = 1
Duration of Immunity
Dafilis M, Frascoli F, et al. (2014) Theor Biol Med Model 11:43. 48 Dafilis M, Frascoli F, et al. (2014) Jour Theor Biol 361, 124-132.
Understanding the Infectious Disease Landscape Collinson MS, Khan K, Heffernan JM (2015), PLoS One, 10(11) e0141423. Li X, Jankowski HJ, et al. (2014) BIOMAT 2015, proceedings. Li X, Jankowski HJ, et al. (2014) BIOMAT 2014, proceedings. Laskowski M, Dubey P, et al. (2014) BIOMAT 2014, proceedings. Collinson MS, Heffernan JM (2014) BMC Public Health 14(1), 376 Richardson K, Sander B, et al. (2014), AIMS Public Health, 1(4), 241-255. Duvvuri VRSK, Moghadas SM, et al. (2010). Influenza and other Respiratory Viruses 4(5), 249-258.
Influenza - clusters Li X, Jankowski HJ, et al. (2014) BIOMAT 2014, accepted
Influenza Clusters Li X, Jankowski HJ, et al. (2014) BIOMAT 2014, accepted
Effects of Mass Media V
vaccination
S0
infection
E
Infected vs time
Waning
S2
S1
I
Social distancing recovery
M
Reports: confirmed symptomatic cases EI
R
Media Reports – 2009 H1N1
Collinson MS, Khan K, Heffernan JM (2015), PLoS One Collinson MS, Heffernan JM (2014) BMC Public Health 14(1), 376
Where Can We Use More Mathematical Modelling? Heffernan JM, Chit A. et al., in prep. Richardson K, Sander B, et al. (2014), AIMS Public Health, 1(4), 241-255.
Warning! Important implications to public health initiatives to identify best population based strategies on the availability of vaccine and antivirals. Individual vs. population health Social behaviour Vaccine uptake, drug therapy
Fast-slow analysis Game theory
Mutation and evolution of resistance Too much to list Seasonal forcing Variability In-host Stochastic models Epidemic
Thank You!
Lindi Wahl (Western) Matt Keeling (Warwick) Jianhong Wu (York) Yijun Lou (Hong Kong) Federico Frascoli (Swinburne) Yan Wang (U Petroleum China) Beni Sahai (Cadham Lab Hossein Zivari Piran (York) Jodie McVernon (Melbourne) James McCaw (Melbourne) Redouane Qesmi (Fez Morroco)
Robert Smith? (Ottawa) Shannon Collinson (MOHLTC) Kamran Khan (St Michael’s Hospital, BlueDot) MI2 Lab members (past/present) Centre for Disease Modelling Waterloo Institute for Complexity and Innovation (WICI)
Questions
Puzzle pieces Overlap Multiscale
56
Appendix
Chronic - HIV Modelling/Understanding Oscillations in viral load and T-cell count Viral blips Frascoli F, Wang Y, et al. (2014). CAMQ, in press. Wang Y, Brauer F, et al. (2014) Jour Math Analysis Appl, 414(2), 514-531 Wang Y, Zhou Y, et al. (2013) Jour Math Biol., 67(4), 901-934. Heffernan JM, Wahl L.M (2006). Jour. Theor. Biol., 243(2), 191-204.
Variability in Viral Load  Stochastic Model, measure SD in viral load T-cells Naïve /mL
Lifetime Distributions
Time (Days)
Pharmacokinetics
Cell count .25 loge CD4 T-cells .22 loge CD4 T-cells Viral load .2-.4 log10 RNA copies .32 log10 RNA copies Heffernan JM, Wahl L.M (2006). Jour. Theor. Biol., 243(2), 191-204.
HBV/HCV Persistence ODEs, two compartments of infection death
death
Infection
yL
xL
• Virus- one compartment • Liver perfusion •~400 to >4000 L of blood goes through liver every day
Liver – liver cells budding
Blood PBMCs
xB
v+w
Infection death
yB death
Qesmi R, ElSaadany S, et al. (2011), SIAM J. Appl. Math. 71, 1509-1530. Qesmi R, Wu J, et al. (2010) Math Biosci. 4(2), 118-25.
HBV/HCV Persistence k – production of virus particles a – death rate of infected cells (k L − aL )( k B − aB ) < 0 (k L − aL )( k B − aB ) > 0
Qesmi R, ElSaadany S, et al. (2011), SIAM J. Appl. Math. 71, 1509-1530. Qesmi R, Wu J, et al. (2010) Math Biosci. 4(2), 118-25.
One Exposure
Multiple Exposures
Multiple Exposures Interested in early stages of infection immediately after exposure to a pathogen Assume individual is exposed to infectious dose ‘c’ Unit of pathogen load needed to produce an infection
Therefore, assume infectious pathogen load will grow, overcoming non-specific immune response
Viral Load When V(t,a) > A New exposures do not increase pathogen load
Assume
Holling functional response type 2 b – number of effective contacts K – adjustable parameter that measures how soon saturation can occur
Between Host Model ď&#x201A;§ SEIR model
In general Highlights Immune boosting induces cyclical behaviour in a model of infectious disease dynamics. Seasonal forcing of transmission also induces cyclical behaviour in this system. The birth rate, waning rate and forcing interact to generate complex dynamics. Periodic cycles in the forced system are related to unforced limit cycle dynamics. The “demographic transition” may lead to new dynamical regimes for certain diseases.