Integrating Large Scale Sequencing into Clinical Practice

Heidi L. Rehm, PhD, FACMGHeidi L. Rehm, PhD, FACMGDirector, Laboratory for Molecular Medicine, PCPGMAssistant Professor of Pathology, BWH, MGH, HMS

Laboratory for Molecular Medicine at PCPGM

15% of testing is from Partners’ patients

85% is from other US and International patients

• CLIA-certified in 2003• LMM offers >150 tests in cardiovascular disease, cancer,

hearing loss, pharmacogenetics and genetic syndromes

• Main focus of testing is large multi-gene panels using sequencing technologies (NGS, Sanger)

• Other technologies include TaqMan, Luminex, allele-specific PCR, MLPA, PNAs, STRs, droplet PCR

• Whole genome sequencing diagnostic service

Ryan Shay

Antonio Puerta

DeCarlo Polk

Joe KennedyJason Collier

HCM SCD in Athletes

d. SCD, 7y

38y

Case Presentation

Cardiac gross exam: asymmetric septal LVHCardiac tissue histology: myofiber disarray

Autopsy

40y

6 y

Normal Deceased

Normal

Deceased

Legend:

MI, 49y MI, 70y

39y

d. SCD, 7y6y 3y

38y

SCD = Sudden Cardiac Death

Case Presentation with Family History

40y

6 y

71y 68y

36y

HCM Family 90

Legend:

= Affected individuals

d. SCD, 49y d. SCD, 70y

39y

d. SCD, 7y6y6y 3y

38yLVHArrhythmia

SCD = Sudden Cardiac Death

LVH = Left Ventricular Hypertrophy

40y

Normal Echo

Normal Echo

Normal Echo

Normal Echo

40y

71y 68y

36y

HCM Family 90

Legend:

= Affected Individuals

+ = E187Q positive genotype

- = E187Q negative genotype

d. SCD, 49y d. SCD, 70y

39y

d. SCD, 7y 7y

Normal ECHO

6y 3y

SCD = Sudden Cardiac Death

LVH = Left Ventricular Hypertrophy

+

+- -

- 38yLVHArrhythmia

E187QTPM1

2004: HCM Panel A (5 genes) $3000

HCM Panel B (3 genes) $1150

PRKAG2 and LAMP2 $1500

GLA $600

2007: HCM CardioChip (11 genes) $3000

2011: Pan Cardiomyopathy Test (51 genes)

HCM (18 genes) $3200

(All Genes $3950)

2013: Genome Sequencing (22,000 genes) $9000

Evolution of Cardiomyopathy Testing at Partners

11 genes$6250

Targeted panels are enlarging and more and more labs are launching exome and genome services.

Genomic testing is improving diagnoses, enabling informed treatments, and defining disease risks.

When do we shift to “genomes for all”?

Genetic Testing is Evolving

Need improved analytical validity of NGS

Improve variant calling accuracy

Indels, CNVs, SVs, repeats

Fill in the holes

Targeted NGS panels miss

a variable amount of content

depending on genes

Exome sequencing misses

5-15% of coding sequence

0

50

100

150 NGS Coverage

Courtesy of Birgit Funke

AGMG NGS Guideline

ACMG (www.acmg.net) > Publications > Laboratory Standards and Guidelines > NGS

A Genetic Sequencing Test is Not One Test

DNA

PCRHundreds of assays per sample

SequencingHundreds of bases per exon

Failed exons/bases

One OtoGenome Test is actually ~380,000 tests with an infinite number of possible results. After the NGS process, Sanger follow-up begins:

Detection of full and partial gene deletions through targeted NGS: VisCap

Lo

g2

ra

tio s

am

ple

/ba

tch

me

dia

nUSH2A heterozygous

exon 10 deletion

All exons, sorted by genome position USH2A exons (3’→5’)

OTOF deletion47 exons

Trevor Pugh

However, deletion analysis is not robust in exome and genome sequencing

WGS Case 1: Nonsyndromic Hearing Loss

• Sept 2010: 2 yr old girl born presents to Genetics

• History of congenital bilateral sensorineural hearing loss

• Mild-moderate “cookie-bite” shaped audiogram

• No other complaints

• Genetic testing for hearing loss (Cx26 and OtoChip): NegativeHowever, >100 loci, many without identified genes, are associated to hereditary hearing loss

WGS Case 1: Nonsyndromic Hearing Loss

• July 2011: Patient enrolled into consented research study

• Obtained blood samples on all 11 family members

• Whole genome sequencing ordered on 3 siblings

WGS on 3 children

Variant Analysis Step 1

• ~4 million sequence variants per child

• ~1,250,000 shared variants among the three siblings

• First considered a recessive model (hom or 2 het variants)

• No cause identified

Variant Analysis Step 2

Genome sequence shared with external company

Proposed otopetrin (OTOP1) as causative gene:

Required for normal formation of otoconia in the inner ear.

Two mutations identified:

p.Arg232Trp

4 bp insertion leading to frameshift

Upon manual examination of read data, frameshift variant was deemed a false positive

Copyright 2013 – Partners HealthCare Incorporated – All Rights Reserved

Validation of WGS Alignment and Variant Calling

Variant type1000G

Variants SensitivityWith concordant

zygosity

SNVs 2762933 2735592 (99.0%) 2730826 (98.8%)

Indels 327474 299300 (91.4%) 285401 (87.2%)

Total 3090407 3034892 (98.2%) 3016227 (97.6%)

Concordance with 1000 Genomes Data

Courtesy of Matt Lebo

Copyright 2013 – Partners HealthCare Incorporated – All Rights Reserved

Validation of WGS Alignment and Variant Calling

Variant Type FP (before thresholds) FP (after thresholds)*

SNVs 20 1

Indels 1 0

Variant Type # FN Sensitivity 95% CI

SNVs 410 0 100% 99.1%-100%

Indels 15 0 100% 79.6%-100%

Specificity

Sensitivity

425 Sanger confirmed variants in 195 genes across 700kb of sequence

*Fisher Strand (FS) < 30 and Quality by Depth (QD) >= 4

Courtesy of Matt Lebo

WGS Case 1: Nonsyndromic Hearing Loss

• January 2012: Mother called to say that the 4 oldest children with hearing loss were now complaining of headaches – could this be related?

• Also relayed that a paternal grandfather was reported to have hearing loss detected at age 19 - described inability to hear fire alarm but the clicking of an automatic seat belt annoyed him

Variant Analysis Step 3

• Consider dominant inheritance with reduced penetrance

• Examined known hearing loss genes (~70) for any single variant

• 32 rare/novel variants found across 31 different genes

• All missense variants with no strong data to support a deleterious

impact

Variant Analysis Step 4: Segregation

• Cost to develop Sanger assays for all 32 variants and test

for presence in 11 family members = $5428

• Examined cost to perform linkage analysis by SNP array:

$386.00 per sample x 11 samples = $4246

• Requested additional sample from grandfather

D3S1278 to D3S2453 = chr3:115,124,154-136,278,257 (3q13.31-22.3, 21 Mb)

Courtesy of Jun Shen

Some genes fail analysis by genome/exome sequencing

STRC

Exome Coverage of 73 Hearing Loss Genes

Analyzed case by new OtoGenome Test

STRC pSTRC

STRC pSTRC

Hom deletion of STRC

pSTRC

pSTRCSTRC Gene

100 kb deletion (43.89 Mb to 43.99 Mb)

STRCPseudogene

STRC

100,000 Base Deletion Identified

WGS Case 1: Nonsyndromic? Hearing Loss

• Called family to return cause of hearing loss

• Mother relayed that headaches in 4 affected children had resolved once it was discovered that all 4 needed glasses

• Question: Could STRC mutations cause vision problems?

WGS Case 1: Deafness Infertility Syndrome

Males with this deletion will be infertile due to deletion of the adjacent CATSPER gene

Males can father children through intracytoplasmic sperm injection (ICSI)

WGS Case 2: Distal Arthrogryposis Type 5

Disease is known to be AD and to occur de novo

No known genes for DA5

Skeletal Spine stiffness, Hunched anteverted shoulders, Pectus excavatum, Limited forearm rotation and

wrist extension, Bilateral club feet, Congenital finger contractures, Long fingers, Absent phalangeal

creases, Poorly formed palmar creases, Camptodactyly, Dimples over large joints

Muscle Decreased muscle mass (especially in lower limbs), Firm muscles

Face Triangular face, Decreased facial expression

Ears Prominent ears

Eyes Ophthalmoplegia, Deep-set eyes, Epicanthal folds, Ptosis, Duane anomaly, Keratoglobus,

Keratoconus, Macular retinal folds, Strabismus, Astigmatism, Abnormal electroretinogram, Abnormal

retinal pigmentation

Clinical features:

Case from of Michael Murray, MD

Case 2: Distal Arthrogryposis Type 5

Two de novo mutations in exonic sequence:

ACSM4 – acyl-CoA synthetase medium-chain family member 4

5 nonsense variants identified in ESP; 1 with 6.4% MAF; 4 occur once

PIEZO2: mechanosensitive ion channel

Great candidate, but how to we prove causality for a novel gene-disease association?

Collaboration with Shamil Sunyeav

Matchmaker System for Exomes/Genomes

o Collaboration with Ada Hamosh (OMIM/Hopkins)

o Enhance PhenoDB to become a matchmaker system for unsolved exomes and genomes

o Entry of structured phenotypes

o Ability to upload candidate genes/variants and vcf files

o Matches require data entry and dual notification of both parties

o Build algorithm to prioritize matches based upon phenotype match strength and phenotype/gene/variant prior likelihood parameters

o Goal to also interface with other systems (e.g. Decipher, LOVD, etc)

A New Paradigm in Clinical Genomics

Traditional Paradigm

New Paradigm

What is the biggest bottleneck in clinical

sequencing?

Variant Assessment and Reporting

Somatic Cancer

~15,000 variants interpreted in patient reports to date

Average22 min25 min120 min

Variant Assessment TypeVariant with no dataVariant with dbSNP/ESP data onlyVariant with publications

HCM Gene Mutations – 3000 cases tested

>500 clinically significant mutations identified

66% of clinically significant mutations are seen in only one family

Number of probands

Nu

mb

er

of v

ari

ants

MYBPC3E258K

MYBPC3MYH7R502WW792fsR663H

Hearing Loss Gene Mutations – 2000 Cases Tested

Number of probands

Nu

mb

er

of v

ari

ants

GJB2

35delG

GJB2

V37IGJB2

M34T

USH2A

2299delG

SLC26A4

81% (423/523) of clinically significant variants have been seen in only one family

Variant Analysis for General Genome Report

3,227,455 Substitutions and 418,331 Indels3,227,455 Substitutions and 418,331 Indels

20,240 CDS/Splice Variants20,240 CDS/Splice Variants

25 “Pathogenic” Variants

25 “Pathogenic” Variants

Published as Disease-Causing

Genes

5 PGx Class 1

5 Variants5 Variants

PharmGKB

2 Pathogenic0 Likely Pathogenic6 VUS4 Likely Benign13 Benign

<1%

616 Rare CDS/Splice Variants616 Rare CDS/Splice Variants

LOF in Disease Associated Genes

2 Pathogenic9 VUS

11 Variants11 Variants

~2 CNVs >1kb~2 CNVs >1kb

0 PathogenicReview evidence for gene-disease association and LOF role

Review evidence for variant pathogenicity

Fetus with US finding: ↑NT

PTPN11 p.Ile309ValPublished as “pathogenic” for

Noonan syndrome

Patient contacted author of paper who said he later found the variant in 7% of AJ controls; now feels the variant is benign

Courtesy Sherri Bale

Noonan Syndrome Case

?

LMMCase

To improve our knowledge of DNA variation will require a massive effort in data sharing

U41 Genomic Resource Grant: A Unified Clinical Genomics

DatabaseTo raise the quality of patient care by:

• Standardizing the annotation and interpretation of genomic variants

• Sharing variant and case level data through a centralized database for clinical and research use

• Implementing an evidence-based expert consensus process for curating variant interpretations

We will work jointly with the recipients of the Clinically Relevant Variant Resource awards.

Sequencing Laboratories Which Have Agreed to Share Data

www.ncbi.nlm.nih.gov/clinvar

NIH NCBI ClinVar

ClinVar Contributors (>40)

Group Submissions

OMIM 22954

Partners Healthcare’s Laboratory for Molecular Medicine 7127

International Standards For Cytogenomic Arrays (ISCA) 4557

GeneReviews 2264

ARUP 1415

LabCorp (Correlagen) 1391

Sharing Clinical Report Project – BRCA1/BRCA2 902

GeneDx 589

NHGRI (ClinSeq) 423

Leiden Muscular Dystrophy 114

Emory 43

Total = 41923

LMM’s Disease Area Submissions to ClinVar

Phenotypes Probands Genes Unique Variants

Cardiomyopathy 5485 51 3929

Somatic Cancer 3421 21 178

RASopathies 2781 12 376

Hearing Loss and Related Syndromes 2048 65 2218

Connective Tissue Disorders 915 3 227

Hereditary Cancer 665 9 81

Congenital Heart Defects 91 4 43

Ectodermal Dysplasia 81 1 36

Other 867

Documenting arguments will improve the evidence-based assessment of variants

Current Limitations of ClinVar

Many labs have limited and/or non-

standardized infrastructure to house their

data which limits the quantity, quality and

efficiency of data sharing

GeneInsight LabGeneInsight Lab

Reporting EngineReporting Engine

Laboratory KnowledgebaseLaboratory Knowledgebase

Case RepositoryCase Repository

The GeneInsight Suite

• In clinical use since 2005; over 28,000 cases reported• Commercialized and used by >10 clinical labs/research

consortia to date

Current Limitations of ClinVar

o Genomic data and other case-level data is difficult to share in meaningful ways in the public domain

My lab has 25,000 cases with phenotypic data that I am willing but unable to share through ClinVar right now

o We are gathering stakeholders at NIH in September to discuss policies surrounding case-level data submission and access for NCBI databases including ClinVar and dbGaP

GeneInsight Data Sharing Networks

NetworkedStandaloneDatabases

Bioreference Lab-MGH PathologyGeneInsight Clinical Lab Network

New York Genome Center

Lung Cancer ConsortiumGenome Canada Grant

GeneInsight Lab – View of Networked Data

GeneInsight LabGeneInsight Lab

Reporting EngineReporting Engine

LaboratoryKnowledgebase

LaboratoryKnowledgebase

Case RepositoryCase Repository

GeneInsight ClinicGeneInsight Clinic

Patient Genomic Profile RepositoryPatient Genomic

Profile Repository

Alerting EngineAlerting Engine

EHR ConnectivityEHR Connectivity

Supports Molecular Diagnostic Labs

Supports Molecular Diagnostic Labs

Supports Clinicians WhoOrder Genetic Tests

Supports Clinicians WhoOrder Genetic Tests

The GeneInsight Suite

Keeping up with Genomics

~4% of case per year received medium or high alerts (.33% per month)

GeneInsight Alert Delivery

This screenshot was taken from a demonstration system – the content of this screen should not be used for any clinical purpose

Acknowledgements

The General Genome Report