Genomic epidemiology as translational research:
from basic science to public health

Sidney M. Bell, PhD (@sidneymbell)

UC Santa Cruz - Genomics Institute Seminar - August 2022

Impact of translating basic research to public health is
often faster and broader than translating to the clinic

https://www.hiv.gov/hiv-basics/overview/history/hiv-and-aids-timeline

1984

Warn against injection drug use

Offer sex education

Blood screening

1987

First high-sensitivity tests

First ART available

Virtuous cycle

Data generation
Problem identification / hypothesis generation

Methods development

Scientific findings

Application in public health practice

Genomic epidemiology

Using pathogen genomic sequences to understand the distribution and spread of infectious disease

Why genomic epidemiology?

Proven impact when resources + tools + expertise are co-located
Problem space is largely tool development, capacity building & bridging between disciplines
Aligned with our goals of democratizing methods, fostering interdisciplinary collaborations, and promoting open data

Agenda

Brief introduction to gen epi
Gen epi research
- Dengue antigenic evolution
Gen epi practice
- COVIDTracker (2020 - 2021)
- Tool development (2021 - 2022)
Future directions for the field
Discussion and questions

Viruses evolve and spread on similar timescales

Grubaugh, Nature Micro, 2019

Outbreak spread

Sample some individuals

Determine phylogeny

Agenda

Brief introduction to gen epi
Gen epi research
- Dengue antigenic evolution
Gen epi practice
- COVIDTracker (2020 - 2021)
- Tool development (2021 - 2022)
Future directions for the field
Discussion and questions

Dengue is a mosquito-borne virus

that kills 15,000 children annually

Four (uniform?) serotypes of dengue

DENV1

DENV3

DENV4

DENV2

Each serotype is genetically diverse

DENV1

DENV3

DENV4

DENV2

Lifelong protection

Temporary

cross-protection

Initial

response

Original antigenic sin drives

dengue case outcomes

After

1-3 yrs

Increased risk

Antigenic variation changes what the virus looks like to your immune system

Measles

Antigenically uniform

Flu

Antigenically

diverse

Lifelong protection

Temporary cross-protection

Antigenic variation is

why you need a new flu shot every year

Measles is

antigenically uniform

Flu is

antigenically diverse

DENV antigenic relationships

are poorly understood

Serotypes are genetically distinct

Clades are genetically distinct

Serotypes are antigenically distinct

Are clades antigenically distinct?

≠

Titers approximate pairwise

antigenic distance

Experimentally measure how well sera inhibits viral plaque formation

Sera produced by 1st infection

Test viruses

Dengue serotypes look antigenically diverse

NHP sera; 3 months post primary infection

= 2-fold change in PRNT50

Katzelnick et al, Science 2015

How does

dengue evolve

antigenically?

Dengue serotypes look antigenically diverse

Bell et al, eLife 2019

Antigenic distance between viruses i, j

D_{ij} \propto \sum_{m}d_m

Model antigenic distance as the sum of mutation-specific antigenic effects

Mutations between i,j

Squared training error

L1 regularization
(i.e., most values are 0)

\propto

Mapping genotype to phenotype identifies ~30 mutations with antigenic effects

Bell et al, eLife 2019

(95% CI 0.74–0.77)

(95% CI 0.77–0.79)

RMSE=0.75

R^2=0.78

Each serotype of dengue contains multiple

distinct antigenic phenotypes

Bell et al, eLife 2019

Sanofi vaccine trial efficacy

varies by genotype

DENV4-I

associated with

adverse events

DENV1

DENV3

DENV4

DENV2

Does antigenic diversity impact dengue population dynamics?

Serotypes cycle through populations

Genotypes

Bell et al, eLife 2019

Population susceptibility:

Previously circulating:

Population

immunity:

Clade growth:

Fitness based on antigenic distance from

standing population immunity

Does antigenic novelty contribute

to dengue fitness?

Relative frequency of j

P_i(t) \propto \sum_{j} x_{j}(t) * D_{ij}

Antigenic distance* between i and j

Lukzsa and Lassig, Nature, 2014

Antigenic fitness of i

* serotype level

or genotype level

Predict clade growth

based on antigenic fitness*

Lukzsa and Lassig, Nature, 2014

\propto e^{(P_i(t) + dt)}

Growth rate @

next 5 years

* serotype level

or genotype level

Bell et al, eLife 2019

Fitness based on antigenic distance* from

standing population immunity

Bell et al, eLife 2019

Serotype-level antigenic fitness drives clade growth & decline

Bell et al, eLife 2019

Summary

Dengue virus undergoes ongoing, slow
antigenic evolution within each serotype
Antigenic novelty is a strong determinant
of serotype-level population dynamics

Agenda

Brief introduction to gen epi
Gen epi research
- Dengue antigenic evolution
Gen epi practice
- COVIDTracker (2020 - 2021)
- Tool development (2021 - 2022)
Future directions for the field
Discussion and questions

COVIDTracker

Applied genomic epidemiology for
local public health departments

2020 - 2021 Demonstrate value

2021 - 2022 Work ourselves out of a job

COVIDTracker

May 2020 - June 2021

COVIDTracker program scale

30% of counties in CA
13k sequences publicly released
100+ hours of small-group training

Applied gen epi delivered major local impact

Tracking introductions
County A caught transition from travel-associated (March/April '20) to community-based cases (July/August)
Ambiguous contact tracing
County B jail outbreak, overlaid with facility locations and transfer records
Related outbreaks
Shared genomes across skilled nursing facilities in County C identified shared staff as a major issue
...and dozens more

Work yourself out of a job

Ops

Four pillars of successful, applied gen epi

Building software to enable data management & bioinformatics

PH4GE schema
relational database
Intuitive UI
Contextual data from GISAID
Constantly updated integrations with Nextstrain, Nextclade, USHeR & Pangolin

Context is everything

Linking data is a huge challenge

DOH ID	CZB ID	GISAID ID

Queryable, version-controlled, single source of truth

ACTG...

Scripted, reproducible QC & disambiguation

Linking data is a huge challenge

Building local health departments' capacity for in-house sequencing & interpretation

Alli Black!

Hands-on training for sequencing, hundreds of hours of interpretation guidance

Pilot experiment: can we build software to automate basic interpretation?

Tree traversals to examine topology & patristic distances
Automated narrative interpretations for common epidemiological questions
Visualizations germane to epidemiology
Ability to overlay and filter on arbitrary epidemiological metadata

Agenda

Brief introduction to gen epi
Gen epi research
- Dengue antigenic evolution
Gen epi practice
- COVIDTracker (2020 - 2021)
- Tool development (2021 - 2022)
Future directions for the field
Discussion and questions

Scientific

Topology-based clustering
Inference of sampling density
Viz & interpretation aids for
partially-sampled outbreaks
Actionable metrics for interpreting genomic
surveillance data

Systemic

Data management - linkage between patient data, epi data, and genomic data
Sequencing & operations -
major equity issues
Professional currency for academics doing capacity building & applied public health work

Needs in the field

Acknowledgements

Bedford lab, especially Trevor Bedford, Richard Neher, Alli Black, Gytis Dudas & Leah Katzelnick
NSF GRFP for funding dengue work

CZI + CZB Genomic Epi team, especially Patrick Ayscue, Alli Black, Shannon Axelrod, Tony Tung, Olivia Holmes, Josh Batson, David Dynerman, Amy Kistler, Dan Lu, Jack Kamm, Maira Phelps, Kirsty Ewing, Hana Zaydens, Colin Megill & Phoenix Logan

Questions?

@sidneymbell

https://sidneymbell.science

Genomic epidemiology as translational research: from basic science to public health

Impact of translating basic research to public health is often faster and broader than translating to the clinic

Virtuous cycle

Genomic epidemiology

Using pathogen genomic sequences to understand the distribution and spread of infectious disease

Why genomic epidemiology?

Agenda

Viruses evolve and spread on similar timescales

Outbreak spread

Sample some individuals

Determine phylogeny

Agenda

Dengue is a mosquito-borne virus

that kills 15,000 children annually

Four (uniform?) serotypes of dengue

Each serotype is genetically diverse

Original antigenic sin drives

dengue case outcomes

Antigenic variation changes what the virus looks like to your immune system

Antigenic variation is

why you need a new flu shot every year

DENV antigenic relationships

are poorly understood

Titers approximate pairwise

antigenic distance

Dengue serotypes look antigenically diverse

How does

dengue evolve

antigenically?

Dengue serotypes look antigenically diverse

Model antigenic distance as the sum of mutation-specific antigenic effects

Mapping genotype to phenotype identifies ~30 mutations with antigenic effects

Each serotype of dengue contains multiple

distinct antigenic phenotypes

Sanofi vaccine trial efficacy

varies by genotype

Does antigenic diversity impact dengue population dynamics?

Serotypes cycle through populations

Genotypes

Fitness based on antigenic distance from

standing population immunity

Does antigenic novelty contribute

to dengue fitness?

Predict clade growth

based on antigenic fitness*

Fitness based on antigenic distance* from

standing population immunity

Serotype-level antigenic fitness drives clade growth & decline

Summary

Agenda

COVIDTracker

Applied genomic epidemiology for local public health departments

2020 - 2021 Demonstrate value

2021 - 2022 Work ourselves out of a job

COVIDTracker

COVIDTracker program scale

Applied gen epi delivered major local impact

Work yourself out of a job

Four pillars of successful, applied gen epi

Building software to enable data management & bioinformatics

Context is everything

Linking data is a huge challenge

Linking data is a huge challenge

Building local health departments' capacity for in-house sequencing & interpretation

Pilot experiment: can we build software to automate basic interpretation?

Agenda

Scientific

Systemic

Needs in the field

Acknowledgements

Questions?

@sidneymbell

Genomic epidemiology as translational research:
from basic science to public health

Impact of translating basic research to public health is
often faster and broader than translating to the clinic

Applied genomic epidemiology for
local public health departments