Disclosures Precision Medicine Approaches for Studying Preterm - - PowerPoint PPT Presentation

2018/4/21 1 Precision Medicine Approaches for Studying Preterm Birth: The Role of Big Data

Marina Sirota, PhD

Assistant Professor

Sirota Lab

Disclosures

• Currently scientific advisor at T

woXAR

• Shareholder in Somnics
• Previously a Consultant at RevMed, NGM Pharmaceuticals and

Novartis

• Previously employed at Pfizer
• Invited talks at Genentech, Novartis, 23andme, DNANexus,

Lifecode, RevMed, Circuit Therapeutics

Why Now?

Moore’s Law – Biology and Computation

Cost Per Genome Cost of Computational Resources

2018/4/21 2

Integrative Personal “Omics” Profiling

Genome Transcriptome Epigenome Microbiome Proteome / Metabolome Antibodyome

Pregnancy and Preterm Birth Why Study Preterm Birth?

Worldwide, 15 million babies are born premature each year. One million of these infants die within the first 28 days of life. In nearly half of the cases of premature birth, there is no clear cause.

1st trimester (T1) 2nd trimester (T2) 3rd trimester (T3)

Preterm birth (PTB) - live birth before week 37 of gestation

2018/4/21 3 Integrative Computational Approaches for Preterm Birth Research

• Elucidating Genetic and Environmental Determinants of

Preterm Birth

• Nadav Rappoport, Aolin Wang, Hongtai Huang
• Microbiome Meta-Analysis in Preterm Birth
• Idit Kosti
• Transcriptomic Meta-Analysis in Preterm Birth
• Bianca Vora and Aolin Wang
• Integrating Clinical and Molecular Measurements to Study

Preterm Birth

• Carolyn Wang, Brian Le and Idit Kosti

Integrative Computational Approaches for Preterm Birth Research

• Elucidating Genetic and Environmental Determinants of

Preterm Birth

• Nadav Rappoport, Aolin Wang, Hongtai Huang
• Microbiome Meta-Analysis in Preterm Birth
• Idit Kosti
• Transcriptomic Meta-Analysis in Preterm Birth
• Bianca Vora and Aolin Wang
• Integrating Clinical and Molecular Measurements to Study

Preterm Birth

• Carolyn Wang, Brian Le and Idit Kosti

Evidence for the Genetic Component to Preterm Birth

• Both preterm and postterm births tend to recur in mothers
• Birth timing trends among family members
• T

win studies estimate that ~30% of variation in preterm birth is genetic

• Racial disparities in preterm birth
• Large studies carried out on the maternal side, not much on

the fetal genetics to date

Plunkett and Muglia 2008, Zhang 2017

Genome-wide Association Study Identifies Ancestry Specific Variants Associated with Spontaneous Preterm Birth

Nadav Rappoport*, Jonathan Toung*, Dexter Hadley, Ron Wong, Kazumichi Fujioka, Jason Reuter, Charles W Abbott, Sam Oh, Donglei Hu, Celeste Eng, Scott Huntsman, Dale L Bodian, John E Niederhuber, Xiumei Hong, Ge Zhang, Weronika Sikora- Wohfeld, Christopher R. Gignoux, Hui Wang, John Oehlert, Laura Jelliffe, Jeffrey Gould, Gary Darmstadt, Xiaobin Wang, Carlos Bustamante, Esteban Burchard, Michael Snyder, Elad Ziv, Nikolaos A Patsopoulos, Louis J Muglia, Gary M. Shaw, Hugh

• M. O'Brodovich, David K. Stevenson, Atul J. Butte*, Marina Sirota*

Scientific Reports, 2018

2018/4/21 4

Integrating Genetic Data

• Cases: 1726 PTB individuals*

– 900 Male; 826 Female – Genomic DNA extracted from newborn screening bloodspots – Illumina Human Omni 2.5

• Controls:

– Health and Retirement Study (HRS) is a longitudinal study with genotyping for over the age of 50, ~10K individuals – Illumina Human Omni 2.5

• Population Mapping – 1,000 Genomes

* Wang et al, 2013

Key Insight

• Premature babies are at risk of respiratory distress syndrome (RDS)

because their immature lungs do not produce enough surfactant, a protein that keeps small air sacs in lungs from collapsing

• March of Dimes helped develop surfactant therapy, which was

introduced in 1990

• Prior to 1990, babies born between 24-28 weeks gestation were

highly unlikely to survive and thus are likely underrepresented in adults over 25 years of age HRS participants (55-80+ year olds) can serve as controls against the PTB cohort

Extreme Preterm Birth (PTB) Cases < 30 Weeks Health and Retirement Study (HRS) Controls (12,570 Individuals) Spontaneous PTB (1349 Individuals) Americas (AMR) African (AFR) European (EUR) South Asian (SAS) East Asian (EAS) Population-Specific Case Control Analysis Use 1000 Genomes to map populations

Genetic Associations with PTB in the AMR Cohort

Nadav Rappoport, PhD

2018/4/21 5 Genetic Associations with PTB in the AMR Cohort

2 4 6 8 10

• log10(p−value)

rs1979081 0.2 0.4 0.6 0.8 r2 OR4F21 RPL23AP53 ZNF596 FAM87A FBXO25 TDRP ERICH1 ERICH1−AS1 0.2 0.4 0.6 0.8 Position on chr8 (Mb) Plotted SNPs

Nadav Rappoport, PhD Hongtai Huang, Amy Padula, Tracey Woodruff, PTBi Hongtai Huang, Amy Padula, Tracey Woodruff, PTBi

Inpatient Discharge Data Version 3.0 Database

• 37 Exposure Indicator Variables
• 8,035 Census Tracts
• 58 Counties
• Year 2009-2012
• Preterm Birth Prevalence: 6.84%
• 6,503 Census Tracts with ID labels

Initial Matching Strategy Exact Match by Census Tract Information (40% Matched) Match by Census Tract based on Area Weighting Method (74% Matched) Secondary Matching Strategy ZIP Code Interpolation Method Match by Census ZIP Code Tabulation Area (99% Matched) N=1,797,284 N=1,812,145 Hongtai Huang (Submitted)

2018/4/21 6

Hongtai Huang (Submitted)

Integrative Computational Approaches for Preterm Birth Research

• Elucidating Genetic and Environmental Determinants of

Preterm Birth

• Nadav Rappoport, Aolin Wang, Hongtai Huang
• Microbiome Meta-Analysis in Preterm Birth
• Idit Kosti
• Transcriptomic Meta-Analysis in Preterm Birth
• Bianca Vora and Aolin Wang
• Integrating Clinical and Molecular Measurements to Study

Preterm Birth

• Carolyn Wang, Brian Le and Idit Kosti

Microbiome Meta-Analysis: Challenges and Opportunities

• Impossible to combine processed data, need to go back to the

raw data

• Sample metadata is hard to get access to
• Lots of potential biases:

Stephen Nayfach and Katie Pollard, 2016

– 16s vs. whole genome – Taxa vs. gene analysis – Sample prep differences – Variable regions of the 16s gene that were sequenced – Longitudinal vs. Case-Control

Public Pregnancy Microbiome Data

# of samples Mother/ Baby # of Subjects # of body sites # of Species # of reads Reference Body sites Type of Data Time points HMP 5,298 Controls 242 18 5385 30,336,944 The Human Microbiome Project Consortium Mouth, Nose, Skin, Gut and Vagina 16s Controls Stanford PTB Longitudinal (MOD) 4,399 Mother 51 11 4694 656,308 DiGiulio et al. Vagina, Gut, Saliva, and Tooth/Gum 16s, WGS Weekly Placenta Penn 69 Mother 6 5 2108 207,5881 Lauder et al. Placenta, Vagina 16s Delivery Vaginal PTB 349 Mother 100 1 550 2,213,608 Romero et al. Vagina 16s Every 4 weeks until 24, every 2 weeks until the last prenatal visit Placenta PTB 48 Mother 48 1 TBD 1,348,416 Aagard et al. Placenta 16s, WGS Delivery Gut Microbiome in Pregnancy 972 Mother/ Baby 91 1 2339 112,8572 Koren et al. Gut 16s T1,T3,PP Vaginal Pregnancy TBD Mother 42 1 TBD TBD MacIntyre et al. Vagina 16s T1,T2,T3,PP

2018/4/21 7

Statistical Analysis Operational Taxonomical Unit (OTU) Tables UPARSE (Assign Taxonomy) UPARSE (Trimming, Clustering, Aligning) Raw Reads, Metadata

Microbiome Meta-Analysis Pipeline

Study Name Raw Reads Size HMP 19.03GB Stanford PTB Longitudinal (MOD) 8.98GB Placenta Penn 991.2MB Vaginal PTB 1.28GB Placenta PTB 490.9MB Gut Microbiome in Pregnancy 427.2MB

Is time consuming and computationally intensive

Idit Kosti, PhD >10,000 samples 12,427s of OTUs ~1000 OTUs post filtering

PRELIMINARY

Idit Kosti, PhD Microbiome Meta-Analysis Across Body Sites

Can we identify new microbial species associated with PTB in a longitudinal analysis leveraging public data?

Bacterial Vaginosis is a Known Risk Factor for PTB

• Bacterial vaginosis is a state of an overgrowth of anaerobic

bacteria, replacing the normal vaginal Lactobacillus.

• Bacterial vaginosis has been shown to increase the risk for

preterm birth.

• Several microbiome studies have been carried out but no

meta-analysis to date.

Manns-James L., J Midwifery Womens Health, 2011

2018/4/21 8 Meta-Analysis Allows Greater Power

Callahan et al. DiGiulio et al. Hyman et al. Romero et al. Stout et al Total Number of Patients 135 37 82 87 74 PTB Patients 50 5 16 18 23 T1 Samples 42 (10 PTB) 21 (4 PTB) 37 (5 PTB) 6 (2 PTB) 14 (4 PTB) T2 Samples 135 (50 PTB) 31 (4 PTB) 50 (10 PTB) 76 (17 PTB) 55 (18 PTB) T3 Samples 123 (39 PTB) 36 (4 PTB) 46 (9 PTB) 60 (5 PTB) 59 (17 PTB) Overall PTB Ratio 37% 12.5% 21% 17.33% 31.1% Sampling Time Points Once per week Once per week One per trimester Once every 4 weeks (< 24 gestation weeks) Once every 2 weeks (> 24 gestation weeks) Once per trimester

Over 3,000 samples and 350+ women

Why Meta Analysis? Better Balance Between Groups

Idit Kosti, PhD DiGiulio Romero Hyman Callahan Stout Combined Meta-Analysis

Raw 16S Sequencing Data …ACATCATACAGATACAAATA… …ACCCATGATAGAGAAACAGA… HMP 447 samples 386 patients DiGiulio et al. 631 samples 37 patients Romero et al. 323 samples 87 patients Hyman et al. 104 samples 81 patients Data Processing Pipeline UPARSE

Reads preparation, Reads De-replication, Alignment, Taxonomy Prediction, OTU Generation, Tree Generation

OTU Modeling and Meta-Analysis Weighted Linear Mixed Effects Regression

Longitudinal Modeling Accounting for Race, Study, Sampling Frequency

Diversity Analysis

Variance, Shannon Index

Association Analysis

OTU Abundance

Non-Pregnant vs. Term vs. Preterm Samples OTUs Data Normalization

Batch effect Adjustments Accounting for Study Bias

OTU - a microbial taxonomic unit based on sequence divergence.

Idit Kosti, PhD

Stout et al. 131 samples 74 patients Callahan et al. 2012 samples 135 patients

Higher Diversity is Associated with PTB in the First Trimester (T1)

*** *** 0.0050 0.0075 0.0100 10 20 30

GestWeekColl predict(fit_full) Term

Term Preterm

Weeks of Gestation Diversity (Variance) p-value<2.22e-16 1 2 3 Trimester of Collection Diversity (Variance) Idit Kosti, PhD

2018/4/21 9

Linear Mixed Modeling

Trimester 1 Trimester 3 Trimester 2

Romero DiGiulio Hyman

OTU Abundance ~ Trimester of Collection*Outcome + Race + (1|SubjectID) + (1|Source)

random effects White Black Asian Other 1,2,3 Term, Preterm

Sampling Scheme Idit Kosti in collaboration with Svetlana Lyalina and Katie Pollard from the Gladstone Institutes UCSF.

Specific Microbial Genera are Associated with PTB

OTUs

Lactobacillus T1 T2 T3 Prevotella Atopobium Olsenella Gardnerella Dialister Aerococcus Clostridium Sensu Stricto Megasphaera

Idit Kosti, PhD

Summary

• We observe higher microbial diversity in women who deliver

preterm, especially in the first trimester.

• We identify novel bacteria associated with preterm birth using

a meta-analysis approach leveraging public data.

• First trimester vaginal sampling may help identify those at risk

for preterm birth.

Kosti I, Lyalina S, … Pollard K, Butte AJ, Sirota M. Meta-Analysis of Vaginal Microbiome Data Provides New Insights In Preterm Birth. In Preparation.

Integrative Computational Approaches for Preterm Birth Research

• Elucidating Genetic and Environmental Determinants of

Preterm Birth

• Nadav Rappoport, Aolin Wang, Hongtai Huang
• Microbiome Meta-Analysis for Preterm Birth
• Idit Kosti
• Transcriptomic Meta-Analysis in Preterm Birth
• Bianca Vora and Aolin Wang
• Integrating Clinical and Molecular Measurements to Study

Preterm Birth

• Carolyn Wang, Brian Le and Idit Kosti

2018/4/21 10

FDR < 0.1 abs(FC) > 1.3 210 genes

Is There a Common Transcriptomic Signature of PTB?

Before Norm After Norm Removing Outliers Cases and Controls are Matched by Sampling Time Preterm Term Gestational Age at Sampling

p=0.125

339 Samples!!!

Bianca Vora and Aolin Wang, PhD

Maternal Transcriptomic Signature of PTB

FDR < 0.1 abs(FC) > 1.3 210 genes

210 genes: 145 Downregulated 65 Upregulated 18 are Secreted 15 Appear in an Earlier Timepoint Potential Biomarkers!

Patients (Maternal Blood) Genes Bianca Vora and Aolin Wang, PhD (In Press)

• FDR < 0.1
• abs(FC) > 1.3

210 genes

Network Analysis

Downregulated Upregulated

Bianca Vora and Aolin Wang, PhD (In Press)

Antigen receptor-mediated signaling pathway Leukocyte activation Lymphocyte activation T cell activation Leukocyte cell cell adhesion

Innate (Non-Specific) Immune System is Upregulated, Adaptive (Specific) Immune Response is Downregulated

Bianca Vora and Aolin Wang, PhD (In Press)

2018/4/21 11

Specific Genes / Proteins That Are Secreted

Genes FC_GSE46510 FC_GSE59491 FC_GSE73685 Directionality P.Value adj.P.Val HPSE 1.12502566 1.029755132 1.302481986 Upregulated 0.01472767 0.072407187 NLRP3 1.120388112 1.045936747 1.463420748 Upregulated 0.008637567 0.055169509 LRG1 1.304600844 1.012782905 1.266657092 Upregulated 0.001694397 0.02445227 CLC 0.77816604 0.925657768 0.460056969 Downregulated 0.013608709 0.06964973 DPP4 0.906503094 0.939578805 0.7254311 Downregulated 0.007598762 0.051200829 IL1R1 1.088237667 1.085324456 1.324171306 Upregulated 0.006616252 0.047794152 IL1RAP 1.163741033 1.016218763 1.310389926 Upregulated 0.017000877 0.07821447 LAMB2 1.05178463 1.022848768 1.305601216 Upregulated 0.004096624 0.037683905 NELL2 0.785324082 0.91373226 0.642718858 Downregulated 0.000198362 0.010174565 NUCB2 0.749613912 0.965573738 0.800992598 Downregulated 0.000657135 0.01646351 SERPINI1 0.920203336 0.872420873 0.755891687 Downregulated 0.000180817 0.009596761 TFPI 1.044097212 1.117168095 0.750437685 Upregulated 0.009325712 0.057394342 IL1R2 1.231921476 1.051452709 1.625438109 Upregulated 0.004772589 0.040619724 CST7 1.188969828 1.033458464 1.334064851 Upregulated 0.003500353 0.034868265 CD8B 0.856761369 0.909815767 0.742642806 Downregulated 0.006032779 0.045913205

IL1R1 - Interleukin 1 receptor type 1

• Encodes a cytokine receptor that belongs to the interleukin-1

receptor family and is an important mediator involved in many cytokine-induced immune and inflammatory responses

• Associated with several diseases (autoimmune, infection response)
• Investigated in endometrial tissues and chorioamnionitis (PMID:

22572995,15723707)

• IL-1 receptors 1 and 2 and accessory proteins abundance in

pregnant rat uterus at term - regulation by progesterone. (PMID 27440742, 28148737)

TFPI - Tissue Factor Pathway Inhibitor

• Encodes a Kunitz-type serine protease inhibitor that regulates

the tissue factor (TF)-dependent pathway of blood coagulation

• The product of this gene inhibits the activated factor X and

VIIa-TF proteases in an autoregulatory loop

• Elevated plasma TFPI activity causes attenuated TF-dependent

thrombin generation in early onset preeclampsia, associated with placental development (PUBMED 28569919, 28521572, 27566697)

• FDR < 0.1
• abs(FC) > 1.3

210 genes

Is There Predictive Value?

Ishan Paranjpe

Applied an Elastic Net Model with 5 fold Cross Validation Training Set Test Set

2018/4/21 12

Fetal Transcriptomic Signature of PTB

• FDR < 0.1
• abs(FC) > 1.3

210 genes

473 genes: 308 Downregulated 165 Upregulated Innate Immune System is Upregulated

Patients (Cord Blood) Genes Bianca Vora and Aolin Wang, PhD (In Press)

• FDR < 0.1
• abs(FC) > 1.3

210 genes

Maternal and Fetal Signals are Reversed!

Significant Enrichment in Overlap with Maternal Signal p=0.0004

Bianca Vora and Aolin Wang, PhD (In Press)

TLR5 -T

• ll Like Receptor 5
• TLR5 plays a fundamental role in pathogen recognition and activation of

innate immune responses

• Recognizes bacterial flagellin, the principal component of bacterial flagella

and a virulence factor.

• The activation of this receptor mobilizes the nuclear factor NF-kappaB,

which in turn activates a host of inflammatory-related target genes

• TLR2, TLR3 and TLR5 regulation of pro-inflammatory and pro-labour

mediators in human primary myometrial cells (28844021)

• Variants in TLRs associated with PTB and BPD (23047423, 22058078)
• Placental TLR/NLR expression signatures are altered with gestational age

and inflammation (27440318)

Uterine contractions

Influences:

• Infection/Dysbiosis
• Stress/Environmental Factors
• Genetics
• Fetal Congenital Anomalies

Effector molecules (maternal, fetal)

Circulating Maternal Immune System

Adaptive immune cells Innate immune cells Infection

Normal Pregnancy Preterm Labor

Omics readouts:

• Cell-free DNA/RNA (Maternal and Fetal plasma)
• Immune repertoire (BCR & TCR Sequencing)
• Immune phenotyping-flow cytometry, CyTOF
• Microbiome
• Metabolomics

Immune Activation Figure Credit: Tippi MacKenzie, MD

2018/4/21 13 Integrative Computational Approaches for Preterm Birth Research

• Elucidating Genetic and Environmental Determinants of

Preterm Birth

• Nadav Rappoport, Aolin Wang, Hongtai Huang
• Microbiome Meta-Analysis for Preterm Birth
• Idit Kosti
• Transcriptomic Meta-Analysis in Preterm Birth
• Bianca Vora and Aolin Wang
• Integrating Clinical and Molecular Measurements to Study

Preterm Birth

• Carolyn Wang, Brian Le and Idit Kosti

UCSF Electronic Medical Records (EMR)

• Time span: 2012 – today
• Number of patients: 922,596
• Data included:

– Allergies – Diagnosis – Encounters – Immunizations – Lab tests – Medications orders – Procedure orders – Vitals – Imaging

Can We Use EMR Data to Study Preterm Birth?

• WHY?

– Outcomes, outcomes, outcomes! – 9 month time frame – Consistent care

• Carolyn Wang

– Rising Senior at Gunn High School

• Work done under the mentorship of Idit Kosti, PhD
• Oral Presentation at AMIA 2017, Washington DC!

Leveraging Clinical Data to Study Preterm Birth

UCSF Electronic Health Records 2011-2016 1,300 preterm births 12,000 term births

Carolyn Wang and Idit Kosti, PhD

Lab tests Medications Demographics BMI Diagnoses Can we identify those at risk for PTB?

2018/4/21 14

• FDR < 0.1
• abs(FC) > 1.3

210 genes

Building a Trajectory of Pregnancy

Carolyn Wang and Idit Kosti, PhD

Experimental Design – Data Filtering

Carolyn Wang and Idit Kosti, PhD. Total Records Preterm 1,394 | Term 13,997 Established Timeline Preterm 1,039 | Term 9,529 Preterm 977 | Term 8,903 Removed: Start date – End date < 0 Male patients Age < 15 and Age > 60 Preterm 871 | Term 8,771 Removed: Twins and triplets pregnancies Preterm 607| Term 6872 Removed: Start date – End date < 80 Start date – End date > 301 2

• FDR < 0.1
• abs(FC) > 1.3

210 genes

Preliminary Associations

Carolyn Wang and Idit Kosti, PhD Fisher test Other pvalue = 7.72e-03 Black pvalue = 1.63e-05 Asian pvalue = 3.86e-08

Future Directions

• Build longitudinal models to predict pregnancy outcomes
• Examine combinatorial features
• Leverage imaging data (ultrasound) and deep learning

approaches to identify features associated with PTB

• Integrate imaging and clinical datasets to study this complex

phenotype

Slide 54 2 need MUCH BIGGER FONT

Marina Sirota, 9/6/2017

2018/4/21 15

Big Data Precision Medicine

responders non-responders test

Creating a PTB Data Repository

• Funded by the March of Dimes
• Serve as a central data repository for omics data across the 5

MOD Transdisciplenary Centers

• Enabling new scientific questions
• Enhance collaboration and coordination
• Accelerate the pace of discovery

Maya Zuhl, Elizabeth Thomson, Cris Thomas Link: http://www.immport.org/resources/mod

Launched a New Site in Nov 2017!

My Group Aolin Wang Bianca Vora Idit Kosti Silvia Pineda Carolyn Wang Dmitry Rychkov Hongtai Huang Ali Taubes Leandro Lima Kat Yu Akshay Ravoor Brian Le Ishan Paranjpe Manish Paranjpe Shan Andrews Atul Butte Nadav Rappoport Sandra Andorf Sanchita Bhattacharya Hyojung Paik Katie Pollard Svetlana Lyalina Tippi MacKenzie Tracey Woodruff Amy Padula Laura Jelliffe Gary Shaw David Stevenson Dan DiGiulio David Relman Ben Callahan Ron Wong Cele Quaintance Steve Quake Wenying Pan Lance Martin Rebecca Baer

Acknowledgements

Funding Sources NLM, NIA, NIAMS, NICHD, March of Dimes, PTBi, Pfizer, BWF

Tech and Admin Support: Boris Oskotsky, Edna Rodas, Mary Lyall

MOD TDC Sam Parry Rita Leite Aubrey Bailey Louis Muglia Ge Zhang Joanne Chappell Rebecca Anderson Jessica Chubiz Sarah England Neal Sondheimer Antonis Rokas Thuy Ngo Susan Holmes Nima Aghaepour Rob Tibshirani Martin Angst Brice Gaudilliere John Oehlert NG Elizabeth Thomson Jeffrey Wiser Patrick Dunn Cris Thomas Maya Zhul

2018/4/21 16

Thanks!

Thanks! marina.sirota@ucsf.edu http://sirotalab.ucsf.edu We are hiring!