building capacity for automated phylodynamic analysis · 8/1/2017 · building capacity for...
TRANSCRIPT
![Page 1: Building Capacity for Automated Phylodynamic Analysis · 8/1/2017 · Building Capacity for Automated Phylodynamic Analysis A look at the 2015 HIV Outbreak in rural Indiana ... community](https://reader033.vdocuments.mx/reader033/viewer/2022060317/5f0c64957e708231d4352d90/html5/thumbnails/1.jpg)
Building Capacity for Automated Phylodynamic AnalysisA look at the 2015 HIV Outbreak in rural Indiana
Ells Campbell, MSComputational BiologistDivision of HIV/AIDS PreventionNational Center for HIV/AIDS, Viral Hepatitis, STDs, and Tuberculosis
Prevention
CDC Health Information Innovation Consortium
August 1st, 2017
![Page 2: Building Capacity for Automated Phylodynamic Analysis · 8/1/2017 · Building Capacity for Automated Phylodynamic Analysis A look at the 2015 HIV Outbreak in rural Indiana ... community](https://reader033.vdocuments.mx/reader033/viewer/2022060317/5f0c64957e708231d4352d90/html5/thumbnails/2.jpg)
Program Perspective – Data Silos
Laboratory EpidemiologySurveillance
![Page 3: Building Capacity for Automated Phylodynamic Analysis · 8/1/2017 · Building Capacity for Automated Phylodynamic Analysis A look at the 2015 HIV Outbreak in rural Indiana ... community](https://reader033.vdocuments.mx/reader033/viewer/2022060317/5f0c64957e708231d4352d90/html5/thumbnails/3.jpg)
Program Perspective – Data Lake
Surveillance Laboratory Epidemiology
![Page 4: Building Capacity for Automated Phylodynamic Analysis · 8/1/2017 · Building Capacity for Automated Phylodynamic Analysis A look at the 2015 HIV Outbreak in rural Indiana ... community](https://reader033.vdocuments.mx/reader033/viewer/2022060317/5f0c64957e708231d4352d90/html5/thumbnails/4.jpg)
Connecting the Dots
Incidence Surveillance Molecular Surveillance
![Page 5: Building Capacity for Automated Phylodynamic Analysis · 8/1/2017 · Building Capacity for Automated Phylodynamic Analysis A look at the 2015 HIV Outbreak in rural Indiana ... community](https://reader033.vdocuments.mx/reader033/viewer/2022060317/5f0c64957e708231d4352d90/html5/thumbnails/5.jpg)
HIV-TRACE – Quantifying difference between sequences
HIV Sequences in FASTA file
![Page 6: Building Capacity for Automated Phylodynamic Analysis · 8/1/2017 · Building Capacity for Automated Phylodynamic Analysis A look at the 2015 HIV Outbreak in rural Indiana ... community](https://reader033.vdocuments.mx/reader033/viewer/2022060317/5f0c64957e708231d4352d90/html5/thumbnails/6.jpg)
HIV-TRACE Results – Simple Example
A B
D
0.1%
1.4%
![Page 7: Building Capacity for Automated Phylodynamic Analysis · 8/1/2017 · Building Capacity for Automated Phylodynamic Analysis A look at the 2015 HIV Outbreak in rural Indiana ... community](https://reader033.vdocuments.mx/reader033/viewer/2022060317/5f0c64957e708231d4352d90/html5/thumbnails/7.jpg)
Why build transmission networks?
Results directly comparable to traditional contact tracing data• Can readily compare genetic distance and contact tracing
Fine-grain resolution of transmission dynamics• Inform outbreak-specific intervention efforts• Inform future prevention strategies
Identification of unreported direct or indirect transmission links• Aid for partner services and outbreak investigations
![Page 8: Building Capacity for Automated Phylodynamic Analysis · 8/1/2017 · Building Capacity for Automated Phylodynamic Analysis A look at the 2015 HIV Outbreak in rural Indiana ... community](https://reader033.vdocuments.mx/reader033/viewer/2022060317/5f0c64957e708231d4352d90/html5/thumbnails/8.jpg)
Complementary Perspectives
High-Risk Sexual
Network
Epidemiological
![Page 9: Building Capacity for Automated Phylodynamic Analysis · 8/1/2017 · Building Capacity for Automated Phylodynamic Analysis A look at the 2015 HIV Outbreak in rural Indiana ... community](https://reader033.vdocuments.mx/reader033/viewer/2022060317/5f0c64957e708231d4352d90/html5/thumbnails/9.jpg)
Complementary Perspectives
High-Risk Sexual
Network
HIV Genetic Distance Network
Epidemiological Laboratory
![Page 10: Building Capacity for Automated Phylodynamic Analysis · 8/1/2017 · Building Capacity for Automated Phylodynamic Analysis A look at the 2015 HIV Outbreak in rural Indiana ... community](https://reader033.vdocuments.mx/reader033/viewer/2022060317/5f0c64957e708231d4352d90/html5/thumbnails/10.jpg)
Complementary Perspectives
High-Risk Sexual
Network
HIV Genetic Distance Network
IntegratedEpidemiological Laboratory
Both
![Page 11: Building Capacity for Automated Phylodynamic Analysis · 8/1/2017 · Building Capacity for Automated Phylodynamic Analysis A look at the 2015 HIV Outbreak in rural Indiana ... community](https://reader033.vdocuments.mx/reader033/viewer/2022060317/5f0c64957e708231d4352d90/html5/thumbnails/11.jpg)
Analytics Enrichment
Integration
Both Contact & Genetic Minimum Spanning Tree
(MST)
Informatics
![Page 12: Building Capacity for Automated Phylodynamic Analysis · 8/1/2017 · Building Capacity for Automated Phylodynamic Analysis A look at the 2015 HIV Outbreak in rural Indiana ... community](https://reader033.vdocuments.mx/reader033/viewer/2022060317/5f0c64957e708231d4352d90/html5/thumbnails/12.jpg)
Decision Support
Minimum Spanning Tree
(MST)
Informatics
Recency of HIV Infection
ClinicalIntegration
Both Contact & GeneticEstablishedAcute
![Page 13: Building Capacity for Automated Phylodynamic Analysis · 8/1/2017 · Building Capacity for Automated Phylodynamic Analysis A look at the 2015 HIV Outbreak in rural Indiana ... community](https://reader033.vdocuments.mx/reader033/viewer/2022060317/5f0c64957e708231d4352d90/html5/thumbnails/13.jpg)
Outbreak Response – HIV Outbreak among PWID in rural Indiana
Outbreak Details• >220 HIV+ diagnoses
• >90% inject drugs • 4-15x/day• 1-6 partners/injection
• >90% HCV co-infection
• High prevalence of sex exchange for drugs or money
![Page 14: Building Capacity for Automated Phylodynamic Analysis · 8/1/2017 · Building Capacity for Automated Phylodynamic Analysis A look at the 2015 HIV Outbreak in rural Indiana ... community](https://reader033.vdocuments.mx/reader033/viewer/2022060317/5f0c64957e708231d4352d90/html5/thumbnails/14.jpg)
IDU ≤ 3
IDU ≥ 4
1 partner 2+ partners1 partner 2+ partners
Decision Support - Machine Learning- HIV Outbreak in Rural Indiana, 2015
![Page 15: Building Capacity for Automated Phylodynamic Analysis · 8/1/2017 · Building Capacity for Automated Phylodynamic Analysis A look at the 2015 HIV Outbreak in rural Indiana ... community](https://reader033.vdocuments.mx/reader033/viewer/2022060317/5f0c64957e708231d4352d90/html5/thumbnails/15.jpg)
Outbreak Simulations- HIV Outbreak in Rural Indiana, 2015
![Page 16: Building Capacity for Automated Phylodynamic Analysis · 8/1/2017 · Building Capacity for Automated Phylodynamic Analysis A look at the 2015 HIV Outbreak in rural Indiana ... community](https://reader033.vdocuments.mx/reader033/viewer/2022060317/5f0c64957e708231d4352d90/html5/thumbnails/16.jpg)
Outbreak Response – HIV Outbreak among PWID in rural Indiana
Genetic Distance Networkpolymerase region
1.5% distance threshold
• Most HIV sequences are highly similar, representing rapid and recent transmission
![Page 17: Building Capacity for Automated Phylodynamic Analysis · 8/1/2017 · Building Capacity for Automated Phylodynamic Analysis A look at the 2015 HIV Outbreak in rural Indiana ... community](https://reader033.vdocuments.mx/reader033/viewer/2022060317/5f0c64957e708231d4352d90/html5/thumbnails/17.jpg)
Reported Needle
Partners
Node Size
Group C
Group A
Group B
No Group
Node Color
Genetic Distance Network
0.1% distance threshold
• Pruning links with a lower threshold reveals community structure at cost of historical resolution
• Subgroups may be evidence of biological, social, temporal, and/or geographic factors
Outbreak Response – HIV Outbreak among PWID in rural Indiana
![Page 18: Building Capacity for Automated Phylodynamic Analysis · 8/1/2017 · Building Capacity for Automated Phylodynamic Analysis A look at the 2015 HIV Outbreak in rural Indiana ... community](https://reader033.vdocuments.mx/reader033/viewer/2022060317/5f0c64957e708231d4352d90/html5/thumbnails/18.jpg)
Composite of 100 unique MSTs
Outbreak Response – HIV Outbreak among PWID in rural Indiana
• Minimum spanning trees (MSTs) select the most parsimonious links from the distribution of distances
• There can be many equally true MSTs, so we combine many unique MSTs to illustrate uncertainty
![Page 19: Building Capacity for Automated Phylodynamic Analysis · 8/1/2017 · Building Capacity for Automated Phylodynamic Analysis A look at the 2015 HIV Outbreak in rural Indiana ... community](https://reader033.vdocuments.mx/reader033/viewer/2022060317/5f0c64957e708231d4352d90/html5/thumbnails/19.jpg)
CloseGenetic Links
High-Risk
Contacts
N > 10,000
N = 182
MostlyUninformative
Noise
N > 1,500
Probable Transmission Links
Complementary Perspectives- HIV Outbreak in Rural Indiana, 2015
![Page 20: Building Capacity for Automated Phylodynamic Analysis · 8/1/2017 · Building Capacity for Automated Phylodynamic Analysis A look at the 2015 HIV Outbreak in rural Indiana ... community](https://reader033.vdocuments.mx/reader033/viewer/2022060317/5f0c64957e708231d4352d90/html5/thumbnails/20.jpg)
Genetic Distance Network1.5% distance threshold
Minimum Spanning TreesMean Distance <0.1%
Inferred Transmission Network
Outbreak Response – HIV Outbreak among PWID in rural Indiana
![Page 21: Building Capacity for Automated Phylodynamic Analysis · 8/1/2017 · Building Capacity for Automated Phylodynamic Analysis A look at the 2015 HIV Outbreak in rural Indiana ... community](https://reader033.vdocuments.mx/reader033/viewer/2022060317/5f0c64957e708231d4352d90/html5/thumbnails/21.jpg)
Integration is Complex and Resource Intensive
HIV Surveillance Data Integration Workflow
![Page 22: Building Capacity for Automated Phylodynamic Analysis · 8/1/2017 · Building Capacity for Automated Phylodynamic Analysis A look at the 2015 HIV Outbreak in rural Indiana ... community](https://reader033.vdocuments.mx/reader033/viewer/2022060317/5f0c64957e708231d4352d90/html5/thumbnails/22.jpg)
Data PrepTrifacta Wrangler • Summary statistics
– Interactive visualization1
• Data cleaning– Guided by machine learning2
– Live preview of changes3
• Optimized for collaboration– Shareable scripts4
• Personal & Enterprise versions available
31
2
4
https://www.trifacta.com/start-wrangling/
![Page 23: Building Capacity for Automated Phylodynamic Analysis · 8/1/2017 · Building Capacity for Automated Phylodynamic Analysis A look at the 2015 HIV Outbreak in rural Indiana ... community](https://reader033.vdocuments.mx/reader033/viewer/2022060317/5f0c64957e708231d4352d90/html5/thumbnails/23.jpg)
Exploratory AnalysisAlpine Chorus Powerful collaborative and analytics tools via web forms
Intuitive workflow editor with drag and drop “operators”
Built-in analytics (decision trees, linear regression, PCA, k-means, and dozens more)
Custom code turned into drag & drop operators (R, Python, Java)
Open source and Enterprisegithub.com/alpinedatalabs
![Page 24: Building Capacity for Automated Phylodynamic Analysis · 8/1/2017 · Building Capacity for Automated Phylodynamic Analysis A look at the 2015 HIV Outbreak in rural Indiana ... community](https://reader033.vdocuments.mx/reader033/viewer/2022060317/5f0c64957e708231d4352d90/html5/thumbnails/24.jpg)
Automated Solution
Enables reproducible machine learning analytics
…without the programmer!
Model #1
Model #2
Model #3
![Page 25: Building Capacity for Automated Phylodynamic Analysis · 8/1/2017 · Building Capacity for Automated Phylodynamic Analysis A look at the 2015 HIV Outbreak in rural Indiana ... community](https://reader033.vdocuments.mx/reader033/viewer/2022060317/5f0c64957e708231d4352d90/html5/thumbnails/25.jpg)
Network visualization, exploration,and animation
Dynamic model generation• Person Person• County County
Variables mapped via dropdown• Shape = Risk Factor• Color = Infection Status• Size = # of Reported Partners• Link width = genetic distance
Pattern Recognition
Low-cost ‘lite’ version on the way for public health
Network VisualizationCentrifuge - Enterprise
![Page 26: Building Capacity for Automated Phylodynamic Analysis · 8/1/2017 · Building Capacity for Automated Phylodynamic Analysis A look at the 2015 HIV Outbreak in rural Indiana ... community](https://reader033.vdocuments.mx/reader033/viewer/2022060317/5f0c64957e708231d4352d90/html5/thumbnails/26.jpg)
MicrobeTRACE
Sequences + Epidemiologic Data Interactive Exploration of Transmission Networks
github.com/cdcgov/microbetrace
![Page 27: Building Capacity for Automated Phylodynamic Analysis · 8/1/2017 · Building Capacity for Automated Phylodynamic Analysis A look at the 2015 HIV Outbreak in rural Indiana ... community](https://reader033.vdocuments.mx/reader033/viewer/2022060317/5f0c64957e708231d4352d90/html5/thumbnails/27.jpg)
MicrobeTRACE
Distance Histogram
Alignment Viewer
Maps
Networks
github.com/cdcgov/microbetrace
Tree Viewer
![Page 28: Building Capacity for Automated Phylodynamic Analysis · 8/1/2017 · Building Capacity for Automated Phylodynamic Analysis A look at the 2015 HIV Outbreak in rural Indiana ... community](https://reader033.vdocuments.mx/reader033/viewer/2022060317/5f0c64957e708231d4352d90/html5/thumbnails/28.jpg)
Potential for Enhancement
BusinessPartner with vendors for licensing innovation
• Public health “ROI” doesn’t translate to increased budget
AdminNetwork administration and security must catch up and keep pace
TechnologyDynamic redaction of PII
• Submit necessary data for automated analysis• Rapidly integrate sensitive data locally
![Page 29: Building Capacity for Automated Phylodynamic Analysis · 8/1/2017 · Building Capacity for Automated Phylodynamic Analysis A look at the 2015 HIV Outbreak in rural Indiana ... community](https://reader033.vdocuments.mx/reader033/viewer/2022060317/5f0c64957e708231d4352d90/html5/thumbnails/29.jpg)
On the Shoulders of GiantsIndiana• Indiana State Department of Health
Scott County Health Department• Clark County Health Department• Foundations Family Medicine • Indiana University School of
Medicine, Division of Infectious Diseases
• University of Louisville School of Medicine, Division of Infectious Diseases
• NCHHSTP, Division of HIV/AIDS Prevention NCHHSTP, Division of Viral Hepatitis NCEZID, Division of Scientific Resources, Specimen Management Branch and Biotech Core Facility
• Epidemic Intelligence Service
Surveillance• HIV Incidence and Case Surveillance
Branch (HICSB)– Molecular HIV Surveillance
• State, Territory, Local, and Tribal Public Health Departments
STOP Study• NYC, NC, and SF public health • NCHHSTP, Division of HIV/AIDS
Prevention (DHAP)
Commercial Laboratories• Quest• LabCorp
Software and Support• SciComp, ITSO and OCISO at CDC• Leidos Inc.
– Centrifuge Systems– Trifacta– Alpine Data Labs
• UCSD Viral Evolution Group (VEG)• Stanford Visualization Group • Seattle Interactive Data Lab• Stanford HIV Drug Resistance DB
Team• Los Alamos National Laboratories• Gephi Consortium