28
Newton Institute, Infectious Diseases Dynamics, Aug 21, 2013 1 Inference of epidemiological dynamics from sequence data: application to influenza Cécile Viboud with Martha Nelson, Eddie Holmes, Julia Gog, Bryan Grenfell Fogarty International Center National Institutes of Health Bethesda, MD, USA

Newton Institute, Infectious Diseases Dynamics, Aug 21, 2013 1 Inference of epidemiological dynamics from sequence data: application to influenza Cécile

Embed Size (px)

Citation preview

Page 1: Newton Institute, Infectious Diseases Dynamics, Aug 21, 2013 1 Inference of epidemiological dynamics from sequence data: application to influenza Cécile

Newton Institute, Infectious Diseases Dynamics, Aug 21, 2013 1

Inference of epidemiological dynamics from sequence data: application to influenza

Cécile Viboudwith Martha Nelson, Eddie Holmes, Julia Gog, Bryan Grenfell

Fogarty International Center National Institutes of Health

Bethesda, MD, USA

Page 2: Newton Institute, Infectious Diseases Dynamics, Aug 21, 2013 1 Inference of epidemiological dynamics from sequence data: application to influenza Cécile

2

Outline

• Influenza has a long-history of fitting epidemiologic models to data

• Recent explosion of sequence data makes epidemiological inference possible

• Contrast insights from both types of analyses• Spatial patterns (Pandemic, epidemic)• Temporal patterns (Growth rate, R0, and else)

Page 3: Newton Institute, Infectious Diseases Dynamics, Aug 21, 2013 1 Inference of epidemiological dynamics from sequence data: application to influenza Cécile

3

>11,900 full genomes sequenced to date

• Majority are human influenza A virus

The NIAID/NIH Influenza Genome Sequencing Project

Page 4: Newton Institute, Infectious Diseases Dynamics, Aug 21, 2013 1 Inference of epidemiological dynamics from sequence data: application to influenza Cécile

Evolutionary analysis using BEASTBayesian evolutionary analysis sampling trees

• Platform for integrating sequence, time, spatial data for – Estimating evolutionary rates– Inferring population dynamics (coalescent)– Phylogeography

Exact date of influenza virus sampling is available (allows fine-scale temporal resolution)

Page 5: Newton Institute, Infectious Diseases Dynamics, Aug 21, 2013 1 Inference of epidemiological dynamics from sequence data: application to influenza Cécile

5

Spatial dynamics

Page 6: Newton Institute, Infectious Diseases Dynamics, Aug 21, 2013 1 Inference of epidemiological dynamics from sequence data: application to influenza Cécile

6

• Multiple introductions of virus into New York state in each season• Little persistence of viral lineages between seasons (• No spatial structure within New York State• Antigenic drift is an episodic process and does not seen to occur in New York State

Global NA phylogeny 1997-2005

Local influenza A Virus Evolution: New York State 1997-2005 (413 full-genome sequences)

Nelson et al, Plos Pathogen, 2006

Page 7: Newton Institute, Infectious Diseases Dynamics, Aug 21, 2013 1 Inference of epidemiological dynamics from sequence data: application to influenza Cécile

7

Spatial Diffusion of A/H1N1 in the United States284 full-genomes, 2006-07

• Multiple introductions, no cross-season persistence, no spatial structure

Nelson et al, Plos Pat 2007

• no. clades no. samples

Page 8: Newton Institute, Infectious Diseases Dynamics, Aug 21, 2013 1 Inference of epidemiological dynamics from sequence data: application to influenza Cécile

8

Temporal dynamics of A/H1N1across the US, 2006-07 season

Nelson et al, Plos Pathogens 2007

Page 9: Newton Institute, Infectious Diseases Dynamics, Aug 21, 2013 1 Inference of epidemiological dynamics from sequence data: application to influenza Cécile

9

Hierarchical spread of influenza in the US

R=1.35

R=1.35

R=1.89

Couplingi,j Popi

aiPopjaj/dij

g

Viboud et al, Science, 2006

Model fitted to long-term influenza epidemiological records

Page 10: Newton Institute, Infectious Diseases Dynamics, Aug 21, 2013 1 Inference of epidemiological dynamics from sequence data: application to influenza Cécile

10

Lemey et al., 2009 PLoS Curr

Phylogeographic analysis of 2009 spring pandemic wave

Page 11: Newton Institute, Infectious Diseases Dynamics, Aug 21, 2013 1 Inference of epidemiological dynamics from sequence data: application to influenza Cécile

Epidemiological models of spring 2009 pandemic diffusion

11

Balcan et al., Plos Currents 2009; Bajardi et al Plos One 2011; Hosseini et al Plos One 2010

Page 12: Newton Institute, Infectious Diseases Dynamics, Aug 21, 2013 1 Inference of epidemiological dynamics from sequence data: application to influenza Cécile

12

Tizzoni et al., BMC Med 2012

Epidemiologic models of fall wave of 2009 pandemic

Page 13: Newton Institute, Infectious Diseases Dynamics, Aug 21, 2013 1 Inference of epidemiological dynamics from sequence data: application to influenza Cécile

Diffusion patterns at national scale (3 US locations)

Nelson et al, J Virol 2011

Seasonal fluH1N1pdmSpring 09

H1N1pdmFall 09

Houston

Milwaukee

NY State

Nelson et al., J Virol 2011

Page 14: Newton Institute, Infectious Diseases Dynamics, Aug 21, 2013 1 Inference of epidemiological dynamics from sequence data: application to influenza Cécile

Different spatial structure in spring and fall 2009

Spatially structured co-circulating lineages

One predominant lineage, no spatial structure

Spring 2009 Fall 2009

Nelson et al., J Virol 2011; but Baillie et al, J Virol 2012!

Page 15: Newton Institute, Infectious Diseases Dynamics, Aug 21, 2013 1 Inference of epidemiological dynamics from sequence data: application to influenza Cécile

15

Epidemiologic patterns of fall 2009 pandemic wave

Schools

Pop sizeHumidity

Prior immunity

Distance

Gog et al., unpubl.

Page 16: Newton Institute, Infectious Diseases Dynamics, Aug 21, 2013 1 Inference of epidemiological dynamics from sequence data: application to influenza Cécile

Fall pandemic outbreak at UC. San Diego

16Holmes et al., J Virol 2011

29,000 students55 full genome H1N1pdm

- 24-33 separate introductions

- 7 clusters- No clustering by time, age, gender or geography

Page 17: Newton Institute, Infectious Diseases Dynamics, Aug 21, 2013 1 Inference of epidemiological dynamics from sequence data: application to influenza Cécile

In contrast, much clearer spatial patterns of the influenza virus in swine

0.1

Viral introductions between regions,based on Markov jump counts

Southern source populations

Midwestern sink populations

Nelson et al., PLoS Pathog 2011

3.3

0.4

9.413.1

Page 18: Newton Institute, Infectious Diseases Dynamics, Aug 21, 2013 1 Inference of epidemiological dynamics from sequence data: application to influenza Cécile

Model testing

Model Log Marginal Likelihood

Bayes Factor Comparison

Rates fixed equally -87.08 --

Population of destination -83.24 3.8

Population of origin -108.32 -21.2

Destination x origin -85.45 1.6

Swineflows -80.99 6.1

Nelson et al. PLoS Pathog 2011

Best-fit swine flu model

Page 19: Newton Institute, Infectious Diseases Dynamics, Aug 21, 2013 1 Inference of epidemiological dynamics from sequence data: application to influenza Cécile

19

Influenza spatial spread: insights from sequence data

• Seasonal flu (national):– No persistence over summer– Lots of co-circulating lineages– Hierarchical patterns of spread observed in

epidemiologic data but not in sequence data• sampling ?• role of mixed infections ?

• Pandemic flu:– International pandemic arrival explained by travel

patterns– Conflicting fall wave patterns nationally

Page 20: Newton Institute, Infectious Diseases Dynamics, Aug 21, 2013 1 Inference of epidemiological dynamics from sequence data: application to influenza Cécile

20

Inference of temporal patterns

Page 21: Newton Institute, Infectious Diseases Dynamics, Aug 21, 2013 1 Inference of epidemiological dynamics from sequence data: application to influenza Cécile

Inference of key epidemiological parameters early in a pandemic outbreak: R0, Tg

21

Fraser et al, Science, 2009

Sequence data

TMRCA: Jan-12-09 (Nov-03 to Mar-2)

Epi data

Page 22: Newton Institute, Infectious Diseases Dynamics, Aug 21, 2013 1 Inference of epidemiological dynamics from sequence data: application to influenza Cécile

Influenza seasonality

22

Page 23: Newton Institute, Infectious Diseases Dynamics, Aug 21, 2013 1 Inference of epidemiological dynamics from sequence data: application to influenza Cécile

Tracking population dynamics through time

23

Captures differences in seasonality and viral diversity between regions

Rambaut et al, Nature, 2008; Bahl et al, PNAS, 2012

Page 24: Newton Institute, Infectious Diseases Dynamics, Aug 21, 2013 1 Inference of epidemiological dynamics from sequence data: application to influenza Cécile

24

Strain interactions

Rambaut et al. Nature 2008; Chen et al, J Mol Evol 2008

Page 25: Newton Institute, Infectious Diseases Dynamics, Aug 21, 2013 1 Inference of epidemiological dynamics from sequence data: application to influenza Cécile

Sampling issues: region and viral subtype

25

No. sequences available

Viboud et al. Phil Trans Roy Soc 2013

Page 26: Newton Institute, Infectious Diseases Dynamics, Aug 21, 2013 1 Inference of epidemiological dynamics from sequence data: application to influenza Cécile

26

Stack et al, Interface, 2010

Sampling issues: time

Cannot go back further than the last bottleneckSampling at the end of an epidemic best

Page 27: Newton Institute, Infectious Diseases Dynamics, Aug 21, 2013 1 Inference of epidemiological dynamics from sequence data: application to influenza Cécile

27

De Silva et al, Interface, 2012

Sampling issues: time

Page 28: Newton Institute, Infectious Diseases Dynamics, Aug 21, 2013 1 Inference of epidemiological dynamics from sequence data: application to influenza Cécile

28

Areas for future research

– Sampling– Estimate R from influenza sequence data for

« typical » epidemic season– Explore seasonal drivers and subtype

interactions in viral population size estimates– Other disease systems have clearer spatial

diffusion patterns (swine influenza, West Nile, rabies)

– Movements of hosts vs mutation rate