Whole Genome Sequencing and Machine Learning …...Title Whole Genome Sequencing and Machine...
Transcript of Whole Genome Sequencing and Machine Learning …...Title Whole Genome Sequencing and Machine...
![Page 1: Whole Genome Sequencing and Machine Learning …...Title Whole Genome Sequencing and Machine Learning to Modernize AMR Diagnostics Author PACCARB Created Date 7/18/2019 10:25:47 AM](https://reader030.fdocuments.net/reader030/viewer/2022041118/5f2f9ebee245827c354ec376/html5/thumbnails/1.jpg)
Whole Genome Sequencing and Machine Learning to Modernize AMR Diagnostics
PACCARB – July 10, 2019
JONG LEE, MBADAY ZERO DIAGNOSTICS
CEO & CO-FOUNDER
COMPANY CONFIDENTIAL © COPYRIGHT 2019 DAY ZERO DIAGNOSTICS 1
![Page 2: Whole Genome Sequencing and Machine Learning …...Title Whole Genome Sequencing and Machine Learning to Modernize AMR Diagnostics Author PACCARB Created Date 7/18/2019 10:25:47 AM](https://reader030.fdocuments.net/reader030/viewer/2022041118/5f2f9ebee245827c354ec376/html5/thumbnails/2.jpg)
• Founded in 2016, spin-off from Kwon Lab at MGH
• Based in Boston, MA
• Developing sequencing-based diagnostic for AMR/S direct from clinical samples
• Providing rapid sequencing based services for HAI outbreak control & hospital epidemiology
Day Zero Corporate Overview
COMPANY CONFIDENTIAL © COPYRIGHT 2019 DAY ZERO DIAGNOSTICS 2
Jong Lee, MBACEO
• Harvard & Harvard Business School
• Experienced MedTech exec & consultant
Doug Kwon, MD PhDCSO
• Harvard & NYU
Miriam Huntley, PhDCTO
• MIT & Harvard• Expert in genomics, • Infectious Disease
MD and Research Lab Director
computational biology
![Page 3: Whole Genome Sequencing and Machine Learning …...Title Whole Genome Sequencing and Machine Learning to Modernize AMR Diagnostics Author PACCARB Created Date 7/18/2019 10:25:47 AM](https://reader030.fdocuments.net/reader030/viewer/2022041118/5f2f9ebee245827c354ec376/html5/thumbnails/3.jpg)
Rapid vs. Comprehensive Tradeoff:DZD Will Deliver Both
COMPANY CONFIDENTIAL © COPYRIGHT 2019 DAY ZERO DIAGNOSTICS 3
Com
preh
ensi
ve
ID &
AST
Speed (time to result)
Mod
erat
e ID
, lim
ited
AST
Lim
ited
ID,
no A
ST
48 hours 24 hours 12 hours 6 hours 3 hours
Culture based methods gated by culture growth
Sensor or PCR based methods limited to selected targets
Com
preh
ensi
vene
ss
Automated Culture
Culture + PCR PCR / Molecular Probes
![Page 4: Whole Genome Sequencing and Machine Learning …...Title Whole Genome Sequencing and Machine Learning to Modernize AMR Diagnostics Author PACCARB Created Date 7/18/2019 10:25:47 AM](https://reader030.fdocuments.net/reader030/viewer/2022041118/5f2f9ebee245827c354ec376/html5/thumbnails/4.jpg)
© COPYRIGHT 2019 DAY ZERO DIAGNOSTICS
Our Mission: Diagnose Infections on Day Zero
6 Hours
COMPANY CONFIDENTIAL 4
Clinical Samples (e.g., Whole Blood)
Species ID & AMR/S Profiles
![Page 5: Whole Genome Sequencing and Machine Learning …...Title Whole Genome Sequencing and Machine Learning to Modernize AMR Diagnostics Author PACCARB Created Date 7/18/2019 10:25:47 AM](https://reader030.fdocuments.net/reader030/viewer/2022041118/5f2f9ebee245827c354ec376/html5/thumbnails/5.jpg)
Technology Required to Enable Clinical Use
Whole Genome Sequencing
Blood2Bac™Sample Prep
COMPANY CONFIDENTIAL © COPYRIGHT 2018 DAY ZERO DIAGNOSTICS 5
Keynome® Algorithm
MicrohmDB®
![Page 6: Whole Genome Sequencing and Machine Learning …...Title Whole Genome Sequencing and Machine Learning to Modernize AMR Diagnostics Author PACCARB Created Date 7/18/2019 10:25:47 AM](https://reader030.fdocuments.net/reader030/viewer/2022041118/5f2f9ebee245827c354ec376/html5/thumbnails/6.jpg)
At 1 CFU/mL, must solve:
1. Relative abundance: human DNA outnumbers bacterial DNA by 8-9 orders of magnitude
2. Absolute abundance: there is only 10’s of femtograms of bacterial DNA
3. Amplification inhibitors: Blood and blood collection containers carry amplification inhibitors
Major Challenges to Culture-Free Pathogen Sequencing from Clinical Blood
Required Host DNA/Cellular Reduction
COMPANY CONFIDENTIAL 6© COPYRIGHT 2019 DAY ZERO DIAGNOSTICS
![Page 7: Whole Genome Sequencing and Machine Learning …...Title Whole Genome Sequencing and Machine Learning to Modernize AMR Diagnostics Author PACCARB Created Date 7/18/2019 10:25:47 AM](https://reader030.fdocuments.net/reader030/viewer/2022041118/5f2f9ebee245827c354ec376/html5/thumbnails/7.jpg)
Blood2Bac: Agnostic Detection of Bacteria in Blood Down to 1 cfu/ml
COMPANY CONFIDENTIAL © COPYRIGHT 2019 DAY ZERO DIAGNOSTICS 7
1 00
1 01
1 02
1 03
A c i n e t o b a c t e r
b a u m a n n i i
E s c h e r i c h i a
c o l i
K l e b s i e l l a
p n e u m o n i a e
E n t e r o b a c t e r
c l o a c a e
S t a p h y l o c o c c u s
a u r e u s
S p i k e d B a c t e r i a M B s / H u m a n M B s ( R a t i o ) 0 2 5 5 0 7 5 1 0 0
A c i n e t o b a c t e r
b a u m a n n i i
E s c h e r i c h i a
c o l i
K l e b s i e l l a
p n e u m o n i a e
E n t e r o b a c t e r
c l o a c a e
S t a p h y l o c o c c u s
a u r e u s
P e r c e n t G e n o m e C o v e r e d
Ratio of bacterial DNA / human DNA reads from 1 CFU
Genome coverage achieved
Target
![Page 8: Whole Genome Sequencing and Machine Learning …...Title Whole Genome Sequencing and Machine Learning to Modernize AMR Diagnostics Author PACCARB Created Date 7/18/2019 10:25:47 AM](https://reader030.fdocuments.net/reader030/viewer/2022041118/5f2f9ebee245827c354ec376/html5/thumbnails/8.jpg)
SEQUENCING COST PER MB OF DATA DROPPING DRAMATICALLY
NEW GENERATIONS ENABLE RAPID, SINGLE SAMPLE SEQUENCING
Thesis: Sequencing Will Be A Diagnostic Utility
COMPANY CONFIDENTIAL © COPYRIGHT 2017 DAY ZERO DIAGNOSTICS 8
$0.01
$0.10
$1.00
$10.00
$100.00
$1,000.00
$10,000.00
2001 2004 2007 2009 2012 2015
![Page 9: Whole Genome Sequencing and Machine Learning …...Title Whole Genome Sequencing and Machine Learning to Modernize AMR Diagnostics Author PACCARB Created Date 7/18/2019 10:25:47 AM](https://reader030.fdocuments.net/reader030/viewer/2022041118/5f2f9ebee245827c354ec376/html5/thumbnails/9.jpg)
WGS Potential to be Highly Comprehensive vs. Biomarker Approach
COMPANY CONFIDENTIAL © COPYRIGHT 2019 DAY ZERO DIAGNOSTICS 9
Top 5 Species: 60%
WGS Potential: 100%Bacterial Species Abundance in Blood Cultures
![Page 10: Whole Genome Sequencing and Machine Learning …...Title Whole Genome Sequencing and Machine Learning to Modernize AMR Diagnostics Author PACCARB Created Date 7/18/2019 10:25:47 AM](https://reader030.fdocuments.net/reader030/viewer/2022041118/5f2f9ebee245827c354ec376/html5/thumbnails/10.jpg)
Traditional AMR Prediction: Resistance Gene Lookup
COMPANY CONFIDENTIAL © COPYRIGHT 2019 DAY ZERO DIAGNOSTICS 10
S. aureus + methicillin K. pneumoniae + cefazolin
• Highly interpretable, backed by scientific understanding
• Limited to a small subset of validated resistance genes– Not all mechanisms of resistance
are well characterized or known– Complex mechanisms difficult to
characterize with presence / absence of genes
• Not comprehensive enough to predict susceptibility
WGS data from Earle, S. G., et al. (2016). Nature microbiology, 1, 16041.Spades + BLAST of ArgANNOT resistance genes
Percent of Isolates Containing Relevant AMR Genes
![Page 11: Whole Genome Sequencing and Machine Learning …...Title Whole Genome Sequencing and Machine Learning to Modernize AMR Diagnostics Author PACCARB Created Date 7/18/2019 10:25:47 AM](https://reader030.fdocuments.net/reader030/viewer/2022041118/5f2f9ebee245827c354ec376/html5/thumbnails/11.jpg)
Keynome Accuracy Improves With Amount of Training Data in MicrohmDB
COMPANY CONFIDENTIAL © COPYRIGHT 2019 DAY ZERO DIAGNOSTICS 11
• Keynome performance improves with data • Performance curves differ between
species/drug combinations
![Page 12: Whole Genome Sequencing and Machine Learning …...Title Whole Genome Sequencing and Machine Learning to Modernize AMR Diagnostics Author PACCARB Created Date 7/18/2019 10:25:47 AM](https://reader030.fdocuments.net/reader030/viewer/2022041118/5f2f9ebee245827c354ec376/html5/thumbnails/12.jpg)
MicrohmDB: Large Scale Dataset of Pathogen Genomes and AMR Profiles
COMPANY CONFIDENTIAL © COPYRIGHT 2019 DAY ZERO DIAGNOSTICS 12
45,000 samples collected to date from multiple hospital microbiology labs
High throughput whole genome sequencing (NextSeq, HiSeq) –25,000 thus far
Link genomes with phenotypic AMR dataBioinformatic annotations
Collect Clinical Isolates Sequence Genomes Link AMR Data
![Page 13: Whole Genome Sequencing and Machine Learning …...Title Whole Genome Sequencing and Machine Learning to Modernize AMR Diagnostics Author PACCARB Created Date 7/18/2019 10:25:47 AM](https://reader030.fdocuments.net/reader030/viewer/2022041118/5f2f9ebee245827c354ec376/html5/thumbnails/13.jpg)
DZD Vision: WGS Diagnostics Enable Large Scale Data Opportunity
COMPANY CONFIDENTIAL © COPYRIGHT 2019 DAY ZERO DIAGNOSTICS 13
GENOMIC DATA
Hospital Outbreak Investigation MicrohmDB
Epidemiology
Antibiotic Target Discovery