Real-time tracking of virus evolution

Trevor Bedford (@trvrb)
29 Oct 2015
Evolution Seminar
JF Crow Institute, UW Madison

Slides at


Phylogenies describe history

Phylogenies describe history

Haeckel 1879

Phylogenies describe history

Phylogenies reveal process

Darwin 1859


With source-sink dynamics, one deme is ancestral


Epidemic process

Sample some individuals

Sequence and determine phylogeny

Sequence and determine phylogeny

Middle Eastern MERS-CoV phylogeny

West African Ebola phylogeny

Global influenza phylogeny

Applications of evolutionary analysis for influenza vaccine strain selection and charting outbreak spread


Influenza virion

Flu pandemics caused by host switch events

Influenza B does not have pandemic potential

Phylogenetic trees of different influenza lineages

Antigenic evolution drives viral dynamics

Antigenic evolution in H3N2

Influenza H3N2 vaccine updates

Vaccine strain selection timeline

Antigenic "match" of key importance

Antibodies by vaccination should effectively bind to circulating viruses. This requires:

  1. Identification of antigenically distinct clades of virus
  2. Prediction of clade growth/decline

Hemagglutination inhibition (HI) assays measure binding

Antigenic cartography

H3N2 population in Feb 2014

H3N2 population in Jun 2014

H3N2 population in Oct 2014

H3N2 population in Feb 2015

Resulted in a mismatched 2014-2015 vaccine


Project to provide a real-time view of the evolving influenza population

All in collaboration with Richard Neher

nextflu pipeline

  1. Download all recent HA sequences from GISAID
  2. Filter to remove outliers
  3. Align sequences
  4. More filtering
  5. Build tree
  6. Estimate frequencies
  7. Export JSON for visualization

Predictive models

A simple predictive model estimates the fitness $f$ of virus $i$ as

$$\hat{f}_i = \beta^\mathrm{ep} \, f_i^\mathrm{ep} + \beta^\mathrm{ne} \, f_i^\mathrm{ne}$$

where $f_i^\mathrm{ep}$ measures cross-immunity via substitutions at epitope sites and $f_i^\mathrm{ep}$ measures mutational load via substitutions at non-epitope sites.

Predictive models

Another approach quantifies phylogenetic branching patterns

We're now working to include quantitative predictions of future clade behavior in nextflu

And also including other predictors, like geography:

Global influenza phylogeny

Track geographic transition events

Global influenza circulation shows source-sink dynamics

Evolutionary analyses can inform influenza vaccine strain selection


Epidemic nearly contained, but resulted in >28,000 confirmed cases and >11,000 deaths

Outbreaks are independent spillovers from the animal reservoir

Person-to-person spread in the early West African outbreak

Continued spread through Dec 2014

At epidemic height, geographic spread of particular interest

Rambaut 2015

Later on, tracking transmission clusters of primary importance

Middle East respiratory syndrome coronavirus (MERS-CoV)

Cases concentrated in the Arabian Peninsula with occasional exports

No evidence of epidemic growth, spill-over transmission clusters

Bats ➞ Camels ➞ Humans

Moving forward, genetically-informed outbreak response requires:

  • Rapid sharing of sequence data, genetic context critical
  • Technologies to rapidly conduct phylogenetic inference
  • Technologies to explore genetic relationships and inform epidemiological investigation


Richard Neher (Max Planck Tübingen), Andrew Rambaut (University of Edinburgh), Colin Russell (Cambridge University), Michael Lässig (University of Cologne), Marta Łuksza (Institute for Advanced Study), Gytis Dudas (University of Edinburgh), Pardis Sabeti (Harvard University), Danny Park (Harvard University), Nick Loman (University of Birmingham) Matthew Cotten (Sanger Institute), Paul Kellam (Sanger Institute), WHO Global Influenza Surveillance Network



  • Website:
  • Twitter: @trvrb
  • Slides: