Genomic tracking of SARS-CoV-2 evolution and spread


Trevor Bedford (@trvrb)
Associate Professor, Fred Hutchinson Cancer Research Center
20 Oct 2020
AMD Virtual
Slides at:

Significant fog of war. Genomic approaches offer orthogonal data source to understand the pandemic.

Epidemic process

Sample some individuals

Sequence and determine phylogeny

Sequence and determine phylogeny


  • Genomic epidemiology of SARS-CoV-2
  • Nextstrain platform for real-time phylodynamics

Detection and sequencing of SARS-CoV-2 in January

Jan 11: First five genomes showed that the outbreak was caused by a novel SARS-like coronavirus

Jan 19: First 12 genomes from Wuhan and Bangkok lack genetic diversity

Single introduction into the human population between Nov 15 and Dec 15 and human-to-human epidemic spread from this point forward


Spent the week of Jan 20 alerting public health officials, and since then have aimed to keep up-to-date


Project to conduct real-time genomic epidemiology and evolutionary analysis of emerging epidemics

with Richard Neher, James Hadfield, Emma Hodcroft, Thomas Sibley, John Huddleston, Louise Moncla, Cassia Wagner, Ivan Aksamentov, Moira Zuber, Eli Harkins, Misja Ilcisin, Kairsten Fay, Jover Lee, Allison Black, Miguel Parades, Sidney Bell, Colin Megill

Nextstrain architecture

All code open source at

Two central aims: (1) rapid and flexible phylodynamic analysis and
(2) interactive visualization

Rapid build pipeline for 3000 SARS-CoV-2 genomes (timings are for a laptop)

  • Align with MAFFT (~20 min)
  • Build ML tree with IQTREE (~40 min)
  • Temporally resolve tree and geographic ancestry with TreeTime (~50 min)
  • Total pipeline (~2 hr)

Current data flow for SARS-CoV-2

  1. Labs contribute directly to GISAID (now have >150k full genomes)
  2. Nextstrain pulls a complete dataset from GISAID every 24 hours
  3. This triggers an automatic rebuild on Amazon Web Services
  4. We manually update new lat/longs, etc...
  5. We push this build online to and tweet the update from @nextstrain

We do one update per week day via Seattle and Basel.

Sequencing and data sharing in almost real-time

Figure by Hadfield and Hodcroft using data from GISAID

Dec/Jan: Emergence from Wuhan in ~Nov 2019

Jan/Feb: Spread within China and seeding elsewhere

Feb/Mar: Epidemic spread within North America and Europe

Mar/Apr: Decreasing transmission with social distancing

Epidemic in the USA was introduced from China in late Jan and from Europe during Feb

Once in the US, virus spread rapidly

Single introduction at the beginning of Feb quickly shows up throughout the country

More recently, with ongoing mitigation
and decreased international travel,
regional clades have emerged
More recently, with ongoing mitigation
and decreased international travel,
regional clades have emerged

Sequencing immediately useful for epidemiological understanding, but selection and functional impacts should also be studied

Significant interest in spike mutation D614G

This mutation occurred in the initial European introduction

D614G is prevalent throughout Europe and mixed in US and Australia

D614G is increasing in frequency across states in US and Australia

D614G is increasing in frequency across states in US and Australia

The success of D614G can be explained by either:

  • D614G is more transmissible and has higher $R_0$
  • founder effects and epidemiological confounding

Additional evidence from Ct values of clinical specimens

Sheffield, UK Seattle, USA

Repeated introductions to the UK suggest transmission advantage of D614G

Advancing genomic epidemiology

  • Better methods for large datasets
  • Distinguishing endogenous spread from importations
  • Tying genomic epidemiology together with richer epi data to better understand local transmission
  • Incorporating within-host variation to improve phylogenetic resolution
  • Integrating clinical data to look for mutations that impact clinical outcomes

Using Nextstrain with your own data

Running analyses with Nextstrain

  • Start with the public "ncov" repo on GitHub using this guide:
  • This is designed to run locally or on a cluster and combines data from GISAID with local data
  • This produces an "Auspice JSON" like ncov.json which can be viewed locally or on

Sharing results with Nextstrain

A few example "builds" at

Nextstrain Narratives

  • Narratives are Markdown posts that allow you to pair narrative text to visualization state
  • Made possible through an early decision to embed visualization state in URL
  • Example narrative for SARS-CoV-2 here:

All this relies on rapid and open sharing of pathogen genomic data

All Nextstrain code is entirely open source and intended to be used by the community. We've been working hard on improving documentation at We've also opened a discussion board for questions at


Genomic epi: Data producers from all over the world, GISAID and the Nextstrain team

Bedford Lab: Alli Black, John Huddleston, James Hadfield, Katie Kistler, Louise Moncla, Maya Lewinsohn, Thomas Sibley, Jover Lee, Kairsten Fay, Misja Ilcisin, Cassia Wagner, Miguel Parades, Nicola Müller, Marlin Figgins, Eli Harkins