cser clinical sequencing exploratory research

CSERClinical Sequencing Exploratory

Research

19 May 2014Jim Evans MD, Ph.D

University of North Carolina at Chapel Hill

Accomplishments, Challenges and Opportunities

Green & Guyer. Nature; 470:204.2011

Medical Science = Clinical Care• Basic science is the foundation of medical practice but the

endeavors demand different approaches • Theory alone is insufficient to guide action in clinical practice• The balance between premature translation and inappropriate

delay in translation is difficult to achieve• The stakes are high

– Misapplication regularly results in profound harm• To patients• To society through increased cost• Through missed opportunities

• Medical practice ultimately involves individual values – Which ultimately can’t be gainsaid

• Therapeutically– Reflexic HRT after menopause– Anti-arrhythmics for PVCs– Excessively strict glucose control in

diabetes• In Screening

– Routine use of PSA screening– Whole body MRIs

A Long History of Premature Application of Medical Intervention

The Stakes are High in the Clinical Application of Genomics

• Patients (& families) make serious decisions when we apply genomic knowledge– False positives lead to:

• Unnecessary surgery; years of unnecessary screening• Premature end to diagnostic pursuit, forgoing the true answer

– False negatives lead to:• Forgoing necessary preventive/therapeutic modalities

– Amplified by misclassification of family members as at-risk or not

– Family planning & abortion– The psychological damage of misinformation ? ?

??

?

Our Values Regarding Genetic Information Vary

1. Yes2. No3. I’m not sure

•Alzheimer Disease•Fatal Familial Insomnia•Spinocerebellar Ataxia•Huntington Disease•CADASIL

A few really bad genetic diseases…

We need to understand how heterogeneous values can be accommodated

If you carry a mutation that essentially guarantees that you will develop a severe, unpreventable & untreatable

neurological disease by age 60 would you wish to be told?

Underlying Rationale of CSER &Related Clinical Sequencing Programs

• Unsystematic introduction of new modalities into clinical care generates poor and misleading data regarding its utility & appropriate clinical applications– Shifts costs to an already constrained medical system

• Nothing is free in the world of clinical medicine• A free but inappropriate clinical test is very expensive

– Makes patients de facto research subjects– Privacy concerns are especially resonant in genomics– Undermines systematic collection & sharing of genomic &

phenotypic information - critical to understanding the medical implications of the genome

Lots of sequencing is already being performed; it’s getting really cheap, why not just introduce it into clinical care and learn that way?

To establish best practices for translation and implementation of genomic medicine

CSER is Seeking Data Regarding:• Patient characteristics that signal potential utility of WG/ES, e.g.

– Diagnostic “hit” rates in different clinical populations– When is somatic analysis of tumors most informative for guiding Rx?

• How to analyze large data sets in a clinical environment?– Quality control– Timeliness– Confidentiality– Integration with EMR– etc.

• Special considerations in different populations– e.g. minorities and traditionally underserved groups

• Different values• Different allele frequencies

• How should we deal with the plethora of highly heterogeneous “non-target” data generated when performing genomic sequencing?

Gene discovery is important but incidental to the main goals of CSER

Aggregate CSER Accomplishments• Implemented and integrated genome-scale sequencing into

the clinical arena with successful sequencing, analysis, reporting and return of results

• Designed & implemented a variety of (relatively) efficient analytic systems and workflows– Established numerous “gene lists” to facilitate analysis– Progress towards open-source analytic workflows

• Implementation of facile reporting mechanisms and formats to providers and patients

• Demonstration that WES from FFPE tumors is equivalent to that from frozen samples

Aggregate CSER DataReported 3/12/2014

• Total Enrollment: 1,532– 1,051 Adults (69%) / 481 Children (31%)– 796 Female (52%) / 736 Male (48%)

• Racial/Ethnic breakdown*:– White 74%– African American 7%– Hispanic 8%– Asian 3%– American Indian 2%– Not Reported 7%

• Total Sequenced: 1049 – 232 Tumor-focused / 817 Germline-focused

• 109 Presentations and Posters / 61 Publications

2(2%)

21(26%)

18 (23%)

39(49%)

Cat. 1 (proven clinical utility)

Cat. 2 (potential clinical utility)

Cat. 3 (other cancer genes)

Cat. 4 (all other genes)

HIGHEST category of mutation PER PATIENT

Cat. 1

Cat. 2

Cat. 3

Cat. 4Only mutations in

“non-cancer genes”

Mutations of “potential clinical utility”

Mutations in other“cancer genes”

Tumor WES results

Mutations of “establishedclinical utility”

Return of Results & Clinical Relevance

• Germline analysis of 817 individuals– “Positive” for diagnostic finding 241 (29%)– Incidental Findings per ACMG list 24 (2.9%)

• Incidental Findings by relaxed criteria 46 (5.6%)

Diagnostic Yield by Patient Category

– BLSNHL – 24 cases• 3 positive• 8 possible• 13 negative

– Neuro – 49 cases• 9 positive• 12 possible• 28 negative

– Cancer - 24 cases• 3 positive• 2 possible• 19 negative

– Cardiology – 21 cases• 3 positive• 5 possible• 13 negative

– Retinal - 33 cases• 7 positive• 11 possible• 15 negative

– Dysmorphology – 24 cases• 5 positive• 3 possible• 16 negative

175 cases with overall 41% positive/possible; 59% negative

46%

43%

21%

38%

54%

33%

ELSI–Related Results• A conscious and planned focus on ELSI is part of

CSER’s core mission• Early results in this sphere:

– Informed consent in the context of genome-scale sequencing

• CSER-wide study systematically analyzed the IC documents from the 9 sites & formulated specific recommendations for best practices

– Defining specific challenges related to clinical genomics• Reliable measurement tools

– Needs/opportunities for instrument development• Factors influencing patient understanding & outcomes• Applications of qualitative, quantitative and mixed-methods

research

ELSI–Related Results• Incidental Findings are an active focus of CSER

– Frequency of “medically actionable” IFs: 3-6%– Degree of choice provided to patients has been an issue

• CSER sites have been instrumental in shaping debate, revealing discrepancies between current policies and patient preferences

– Led to shift in ACMG’s policy for ROR

– Beginning to accrue data addressing what patients wish to know regarding “off target” findings

• Majority of patients “want everything” when asked• But even minimal “real world” barriers (e.g. need to make a phone

call) dramatically decreases desired return

• CSER is collecting data on measures of patient satisfaction/distress with regard to both diagnostic findings and off-target results

Future Challenges & Opportunities• CSER has erected an integrated infrastructure to implement

clinical genomic medicine• The central challenge now is to assess whether the application

of genomic medicine actually improves patient outcomes– Health

• Morbidity & mortality– Patient satisfaction, values, and well being– Economics

• Do we save money? • How much do improved outcomes cost?

• Explore outcomes in diverse settings, including ethnic, socioeconomic, clinical status

Outcomes

A role for CSER/NHGRI since the “market” does not focus on outcomes

Outcomes• Assessing genomic outcomes demands new approaches

– Qualitatively different approaches• New cohorts of patients with a range of diagnoses• Comparisons between those with application of various approaches &

those with standard care (e.g. CER)– e.g WES vs. WGS vs. multigene panels vs. combined approaches– Will feed into efforts of IGNITE and EMERGE

– Systematic, quantitative extension of existing cohorts• Large numbers of patients • Longitudinal follow up of those with positive results

– Impact on care, dissemination to family members, etc.• Longitudinal follow up of those with negative results

– Optimal approaches to dynamic reinterpretation as knowledge / technology evolve

• Outcomes measurement must continue to include ELSI-oriented outcomes

• CSER sites have established a variety of inventive analytic workflows for analysis of clinical genomic data

• Need for controlled investigations to determine optimal approaches in order to maximize efficiency– How to best make difficult trade-offs between sensitivity and

specificity• e. g. phenotypically-driven “gene lists” that maximize specificity vs.

“gene agnostic” approach, maximizing sensitivity– Likely to differ by clinical context

» Cancer vs. ID

• Especially important given time & resource constraints of analysis in the clinical setting

Future Needs, Challenges & Opportunities

Analytical Approaches

Analytical Approaches• Adjudication of a variant’s impact is the single most significant

obstacle to the application of genomic medicine• We are evolutionarily predisposed to tolerate false positives &

therefore over-interpret data– The genome is big & thus coincidences are common– Our protocols for adjudication are wholly inadequate– The stakes are high– There rarely exist “gold standards” for adjudication

• Need for statistical approaches like “mutational burden” & thus large numbers

• Planned overlap– ClinGen (aggregation of multiple sources of variation, ranging

from commercial labs to more basic LSS efforts)– CSER (investigating optimal processes and mechanisms for

curation in the clinical setting) – EMERGE (large scale integration with EMRs)

Variants of Uncertain Significance


• Progress will rely on integration and sharing– Need to develop formats that facilitate data sharing amongst

sites/labs/vendors/clinical settings – Incorporate data science to meet the expected problems of scale– While protecting privacy and complying with HIPPA– The market is poorly equipped (indeed, often antithetical) to

integration and sharing data– Well adapted to a UO1 consortium model

• Integration of genomic data with other “omic” data – Where clinically promising

• e.g. RNAseq in tumor analysis

• Integration with training programs

Integration & Sharing

• Expansion of groups sequenced – New diagnostic categories of patients– Need to consider whether genomic approaches can improve health in the

general population– Carrier screening for interested individuals– Targeted sequencing of selected genes in general population that, when mutated, strongly

predispose to severe but preventable disease– Facilitate longitudinal use of genomic information throughout lifespan

• Expansion regarding venues & populations:– Ensure broad diversity of patients– Beyond the academic setting where warranted

• Natural progression from CSER to IGNITE• e.g. coordination with PBRNs

– Investigate cascade impact on family members• We must carefully consider what we mean by “genomic approaches”


Rational & Targeted Expansion

• Focused testing is usually optimal clinically– Unnecessarily broad testing can lead to waste,

false positives and downstream harm with increased cost

– Genomic corollary is determining the proper role of WGS vs. panels (virtual or otherwise)

• CSER has (appropriately) focused on WG/ES – Is now well equipped to also determine when

WG/ES vs. more targeted sequencing is optimal• Comparisons of WGS vs. WES vs. targeted panels in

differing clinical contexts


Clinical Genomics = WGS

We’re Wasting Data• We don’t know the true penetrance or new mutation

prevalence for almost any Mendelian condition– Ascertainment bias has likely over-estimated penetrance and

underestimated new mutation rate for most conditions• Accurate figures will be critical for both clinical and public

health genomic applications• Boot-strap on existing sequencing efforts

– Create registry for reporting of “off target” mutations in critical disease genes to define true prevalence & mutation rates

• Must include ability to obtain:– Personal medical history– Family history – Long-term f//u– Where possible, parental genotypes

• CSER & EMERGE are logical places to start

Thank [email protected]

Spectrum of Genomic Medicine Implementation: Intensity vs. Breadth

NSIGHT

Individual Patient Focus

CSER

Evidence

Generation

IGNITE

eMERGE

UDN

Depth of Patient Characterization

Breadth of Implementation

Testing Multiple Models

System-Wide

Impact

Series10

2

4

6

8

10

12

Breadth of Implementation

Dep

th o

f Ch

arac

teri

zati

onIntensity of Phenotypic/Genomic

Characterization vs. Breadth of Implementation

NSIGHT 1000

Testing Multiple ModelsIndividual Patient Focus

CSER 3000

Evidence GenerationSystem-Wide Impact

IGNITE51,000

eMERGE25,000

UDN1400

Study ID#

Enrolled patients

# Sequen

cedFrozen tissue

FFPE tissue

NOT Sequenc

ed MO (Adult patient) 285 262 246 16 23

TP (Tumor profiling) 52 44 38 6 8

PO (Pediatric patient) 55 53 44 9 2

ES (External study) 8 8 4 4 0

SC (SU2C Prostate) 69 59 58 1 10

TOTAL 469 426 390 36 43

Summary of Clinical Programs of MI-ONCOSEQ

79% informative7% germline finding

57% Informative and Actionable

22%Informative

Only

13%Non-informative

8% Failed Biopsy/tumor content

n=402

Categories of Mutations Identified Through MI-ONCOSEQ


diabetes– Beta-carotene supplementation

• In Screening– Routine use of PSA screening– Whole body MRIs


ELSI–Related Results• The Outcomes & Measures Working Group

identified priority areas for investigation of psychosocial and behavioral impact of integrating sequencing into clinical process– Identified specific challenges related to clinical

genomic sequencing, including:• Reliable measurement tools • Needs/opportunities for instrument development• Factors influencing patient understanding & outcomes• Importance of qualitative, quantitative and mixed-

methods research

Under-sequenced Medically-Relevant Genes

1. KCNQ1: potassium voltage-gated channel, KQT-like subfamily, member 1 Familial atrial fibrillation; Short QT syndrome; Acquired Long QT syndrome

~50% of Exon 1 poorly covered

2. PMS2: postmeiotic segregation increased 2 (S. cerevisiae) Lynch syndrome; colorectal cancer, endometrial cancer

Multiple exons with poor coverage at 3’ end; 5’ portion of last exon

segmental duplications

Results by Diagnostic Categories

• By clinical category

43%

21%

38%

54%54%

PediSeq Project First 39 cases

(prospective and unknown retrospective)

• By clinical category– BLSNHL = 24 cases– Sudden Cardiac = 9 cases– Mitochondrial = 1 case– Intellectual Disability = 5 cases

3 positive (~13%)8 possible (~33%)13 negative (~54%)


0 positive (0%)0 possible (0%)1 negative (100%)

1 positive (20%)1 possible (20%)3 negative (60%)

47 validation (known)Cases – all positive (100%)

46%

Results by Diagnostic CategoriesFirst 145+ cases

• By clinical category– Neuro = 44 cases– Cancer = 24 cases– Cardiology = 12 cases– Retinal = 33 cases– Dysmorphology= 24 cases


3 positive (~13%)2 possible (~8%)19 possible (~79%)


Total = 145 cases25 positive (17%)33 possible (23%)87 negative (60%)



43%

21%

33%

54%54%(40% positive /possible)

Next two slides should be distributed, along with some general CSER information, to Council ahead of meeting so they can review on their own and I don’t need to waste time on it.

Scope of CSER at 9 Sites

• Clinicians Involved: 154• Non-Health Care Professionals: 180• Number of Institutions 26• Peer reviewed publications 56• Non-peer reviewed publications 5• Posters at regional/Ntl/Interntl venues 14• Presentations 95

Lucia – check on publications/presentations for update

Patient Focus of Individual CSER Sites• Baylor

– WES of tumor and germline in childhood cancers• Brigham & Women / Harvard

– Randomized trial of WGS in healthy adults & those with suspected hereditary cardiomyopathy• Children’s Hospital of Philadelphia

– WES and WGS in four cohorts of children with pediatric disease• SNHL, ID, mitochondrial, sudden cardiac death

• Dana Farber / Broad Institute– WES of tumor and germline in assorted cancers

• University of North Carolina at Chapel Hill– Randomized trial of WES in adults and children with wide variety of likely genetic diseases

• Cardiac disease, neurodevelopmental disorders, dymorphology, cancer, ophthalmology diseases

• University of Washington– Randomized trial of tumor and germline WES to compare usual care vs. provision of genomic information in

adults with CRC• Hudson Alpha (new)

– Children with intellectual dysfunction• Kaiser (new)

– Adult couples in context of preconception carrier screening• University of Michigan (new)

– WES of tumors in adults and children with advanced cancer

Somatic tumor sequencingGermline only analysis

Early tumor report data

Category SubjectsI 2/58 (3%)II 14/58 (24%)III 22/58 (38%)IV 57/58 (98%)I or II 16/58 (28%)I, II or III 32/58 (55%)only IV 26/58 (45%)

Proven or potential clinical utility

Any “cancer gene”mutation

Sept 2013: 58 tumor exomes

NUMBER OF SUBJECTS with mutations in each category

MSH6 c.1170delT (p.Phe391fs)BRCA2 c.2857G>T (p.Glu953*)

Lower penetrance in unselected populations?Diagnostic Variants

ABCA4 c.1622T>C (p.Leu541Pro)ABCA4 c.1805G>A (p.Arg602Gln)

Incidental Variants on WES

Case of Interest IChallenges of IFs

New mutation(s)?

Chance?

Case of Interest II

• Subject dxd at age 6 with “hereditary spastic paraplegia”– Confined to crutches and wheelchair– Painful spasticity on daily basis

• At 36 no alternative dx had been found in spite of multiple referrals and evaluations nationally

On WES: GCH1 [p.Arg216*] mutation, consistent with a Diagnosis of dopa-responsive dystonia

• Rx with oral L-dopa often effective – Dramatic response

Occasional Therapeutic Implications

Case of Interest III

• Child with non-syndromic cone-rod dystrophy• Two likely deleterious mutations in BBS9

• c.1255C>G (p.Pro419Ala)• c.2153A>T (p.Glu748Val)

• BBS9 has not been associated with non-syndromic CRD, but CRD is a feature of BBS

• There are other BBS genes associated with non-syndromic retinitis pigmentosa

Expansion of our understanding of phenotypes

Assorted Challenges Highlighted by CSER Sites in Last Update

• Timely analysis– Especially in setting of diseases with poor prognosis (Baylor, Dana Farber)– When Rx/EOL decisions hinge on analysis (Baylor, Dana Farber)

• Disappointment among some participants when not randomized to receive maximal results (BW)

• Construction of user-friendly reports for full range of users (providers to patients)• Curation and signing out variants

– Complex and time consuming (everyone)– Inaccurate public databases of variants (everyone)– Analytic approach (gene lists vs. gene-agnostic approach (CHOP and UNC)

• Difficulty of providing informed consent given inherent clinical time constraints• Difficulty of / need for independent CLIA confirmation• Eventual expansion of phenotypes associated with mutations• Proliferation of available commercial tests of large panels – how to navigate? • Insufficient knowledge about penetrance of mutations, especially in context of

non-target information

Maintaining a Clinical Focus• MPS is (just another) highly complex medical test• Claims that “soon everyone will have their whole genome

sequenced” are typically predicated only upon its low cost– Even if “free”, perceived low cost is an illusion

• Indiscriminate application of medical tests is very expensive– Morbidity/mortality to individuals (e.g. PSA screening)– Expense to individuals & society

• It should be applied as are other medical tests:– When and if the situation warrants– With targeted genomic analysis in most settings

• Targeting could be actual (e.g. capture) or virtual (at the analytic stage)

– WGS/WES useful in selected diagnostic settings and in research

• Think before burdening the system with a flood of extraneous information that begs for misinterpretation

• CSER is generating data to guide its optimal use

Thank You

[email protected]

Venimus, Vidimus, Secuti Sumus

We’re Biologically Predisposed to Misinterpret Genomic Data

• As humans we are evolutionarily predisposed to discern patterns– Even when not real– The genome is so big it guarantees

amazing coincidences• The bar for applying genomically-

derived data must be set very high– We have to be highly cognizant of the

low prior probability of significance of any variant

For possible NHGRI Council meeting talk about CSER. Use the CSER talk I did for the October 2013 meeting that the NIH folks couldn't’t go to b/o government shut down. For local NCGENES advertisement look at the ACMG talk I’m doing in March 2014

Summarize justification for CSER, some of our results at UNC.

Discuss needs, including:Further collection of data to define dxic yields to guide clinicians use of WES

Will need outcome data and the timeline is too short to obtain from initial cycle

Further ELSI data collection regarding return of IF – might give a one-slide overview of where we stand with need for ACMG 56, the ROR controversy and opt- out solution

ClinGen project and adjudication of variants (moon slide) as major need

To figure out penetrance there is a need for registry of off-target mutations in disease genes that allows ascertainment of a. personal medical hx b. family history c. prospective f/u. Lots of efforts going on that are uncovering mutations – need to harness systematically. Doesn’t need it’s own sequencing project if it can be piggy-backed

Need to assess targteted sequencing of genes in general population for possible public health investigation

Describe public health potential with looking for rare highly penetrant mutations.

Real Results Are Better Than Hypothetical Results

• Most important reasons for wanting to learn IF– To get information about family’s possible health risks– Motivation to do things to prevent or delay future health

problems– To contribute to research

• Most important reasons for NOT wanting to learn IF– Too upsetting to find out about risk for a serious health problem– Information would not give definite answers– No way to cure or prevent health problems

• In-person interview during clinic visit– 15/21 indicated they ‘probably’ or ‘definitely’ will ask for at least

some IFs– However, only 6/21 have formally requested any findings to

date via phone call

Key Questions• Somatic analysis of malignant tumors

– How often does such analysis lead to change in Rx?– Improvement of outcome with genomic guidance

• Diagnostic yield with germline analysis– Will provide important guidance to use of genomic sequencing in the clinic– Will likely be different for different patient categories/phenotypes

• Especially important given the time and resource intensive nature of analysis

• Incidental Findings– Frequency with which they arise– Preferences and actual choices made by subjects with regard to such

findings– Impact of receipt on patients and families

CSER Plenary Session Thursday, 10/1013 from 10:30-11:30

Hey Lucia and Brad,I’m starting to think about the plenary talk that I’m giving on Thursday, 10/10, from 10:30 to 11:30.Would you look at the rough outline and see what you think? I want to make sure that I talk about things that will be of interest to everyone.I will need help from you guys getting the aggregate data. I figure since the quarterly reports are due 9/16 I might be able to get you to help me soon after that with aggregation. I’m not a procrastinator and like getting stuff done in advance!Finally, what other things do you think I should talk about? Please feel free to make suggestions; I’d appreciate your input.•An overview of the CSER goals/rationale. I don’t think I’ll provide many details on each group’s individual project, but will likely have a slide or two that summarize the overall approach and patient mix for each site.•Total numbers enrolled•Total numbers analyzed•Demographics including ethnicities, age, sex•Phenotypic/clinical categories of subjects being examined•Diagnostic hit-rates broken down by clinical/phenotypic category

With some degree of granularity with regard to results that are clearly positive vs. possible explanations vs. VUS vs. negative resultsI’ll probably take a slight diversion and briefly discuss different approaches to analysis, which will help me keep the interest of the informatics folks

•Frequency of incidental findings deemed reportableI’m assuming that each site reports this information?

•A bit of the ELSI experience thus far, recognizing that it will be largely anecdotal at this early stage•I’ll likely discuss a few individual cases that illustrate broader principles

e.g. a finding we have of a reportable incidental finding and the interesting family history data obtained upon reporting that back to the subject)our case of the highly therapeutic finding of dopa-responsive dystonia dxother illustrative cases that other groups might have

Thanks!

> Sounds great. How about if we leave the precise time (20 vs 15 minutes) a bit flexible for now and as we start to craft our talks we can refine if needed? For now I'll plan on ~20 minutes. I won't have some of the aggregate data for a couple of weeks so things are still up in the air a bit.>

• Category I mutations– Established clinical utility for the tumor type tested

• Category II mutations– Potential clinical utility– Genes that are members of cancer pathways, gene

families, or functional groups; targets of approved or investigational therapeutic agents

• Category III mutations– In consensus cancer genes (not included in I and II)

• Category IV mutations– All other mutations

ALK - neuroblastoma

MET - neuroblastoma

PHF6 - neuroblastoma

XIAP - neuroblastoma

Early tumor report dataHIGHEST category of mutation PER SUBJECT

2(3%)

14(24%)

16(28%)

26(45%)

Cat. 1 (proven clinical utility)

Cat. 2 (potential clinical utility)

Cat. 3 (other cancer genes)

Cat. 4 (all other genes)


Based on our current data (n=70 tumor exomes), we would estimate that we detect a tumor exome finding that could potentially change pharmacologic treatment of the patient sequenced (with FDA approved drugs or as part of a clinical trial) in the event of tumor recurrence in 25% of pediatric solid tumor cases. The context/caveat to this estimate is that we are sequencing newly-diagnosed pediatric solid tumor patients, who are generally treated initially on or following relatively well-defined protocols, such that the tumor exome data are generally received while the patient is completing their initial standard treatment. Thus in our project the utility of the tumor exome data is being evaluated in the clinical context of recurrent or refractory tumors.Please see attached for several summary slides in case they would be of any use at the meeting.BestWill

Hey,The gene is BBS9, and her number is 141 if you want to look her up. She has a clinical diagnosis of non-syndromic cone-rod dystrophy. She had two variants that were not reported in the literature (but predicted damaging by Polyphen :-) The two rare variants are BBS9 c.1255C>G (p.Pro419Ala) and BBS9 c.2153A>T (p.Glu748Val). The BBS9 gene has not been associated with cone-rod dystrophy (CRD), but of course, CRD is a feature of BBS. Additionally, there are other BBS genes that are associated with non-syndromic retinitis pigmentosa, so we believe her CRD is associated with the two variants identified in her BBS9 genes. We don't have info on phase of the two variants, and we probably won't get the opportunity to study family members.Let me know what additional information you would like on her!

Discuss briefly the inherent variation in how VUS’s / diagnostic findings are reported. Which affects the “hit rate” as well as having important clinical implications. Stress importance of being conservative.

•Definite or Likely (single, clearly deleterious, previously reported pathogenic mutation in AD disease)

•Possible/VUS – right gene with right phenotype with possible variant•VUS – worth reporting but no or minimal claims as to being diagnostic

• This would include a deleterious mutation in a great gene for the phenotype in a recessive disease where you haven't found a second disease mutation. Report b/o ability to recommend focused sequencing / deletion analysis.

Following slides are from CSER Plenary in October 2013 and ACMG Lessons learned in March 2014

CSERClinical Sequencing Exploratory

Research

NHGRI Genome Sequencing Research Network Meeting

Bethesda, MD8 October – 11 October 2013

Jim Evans MD, Ph.DUniversity of North Carolina at Chapel Hill

Green & Guyer. Nature; 470:204.2011

Medical Science = Clinical Care• Medical Science is the foundation of Medical Practice• But the two endeavors demand different approaches • Theory alone is insufficient to guide practice• The stakes are particularly high

– Misapplication can result in real and profound harm• The balance between premature translation and inappropriate

delay in translation is difficult to achieve– Harm to patients– Increased cost– Missed opportunities

• Medical practice ultimately involves individual values – Which ultimately can’t be gainsaid


diabetes• In Screening

– Routine use of PSA screening– Whole body MRIs



• Patients (& families) make serious decisions on the basis of an alleged causative mutation– Years of unnecessary invasive screening or risk-

reducing surgery• Or forgoing such modalities when actually necessary

• Family planning & abortion• Premature end to “diagnostic odyssey”, thus

forgoing the true answer– Forfeiting possible therapies

Genomic Application in the Diagnostic Setting

? ?

??

?


• Inflicting medical care on the healthy requires great caution

• Healthy people arguably have less to gain than sick people • Different relationship between provider & recipient

– The individual isn't typically seeking us out • Benefits are less obvious

– “Great news! You didn't get sick!”• The downsides are easy to see

– All interventions have downsides• The prior probability that any given genomic finding in an

individual is clinically important is low

Genomic Screening of the General Population

We’re Biologically Predisposed to Misinterpret Genomic Data

• As humans we are evolutionarily predisposed to discern patterns– Even when not real– The genome is so big it guarantees

amazing coincidences• The bar for applying genomically-

derived data must be set very high– We have to be highly cognizant of the

low prior probability of significance of any variant

Our Values Regarding Genetic Information Vary

1. Yes2. No3. I’m not sure

•Alzheimer Disease•Fatal Familial Insomnia•Spinocerebellar Ataxia•Huntington Disease•CADASIL

A few really bad genetic diseases…

We need to design clinically oriented approaches by which patient choice & provider’s duties can both be facilitated

If you carry a mutation that essentially guarantees that you will develop a severe, unpreventable & untreatable

neurological disease by age 60 would you wish to be told?

To begin to arrive at best practices for implementing genomic medicine

CSER is Seeking Data Regarding:• What are the patient characteristics that signal potential utility (or

lack thereof) for applying genome-scale sequencing?– Diagnostic “hit” rates in different clinical populations– When is somatic analysis of tumors most informative for Rx?

• What are the best approaches to analysis of data?– In the clinical setting, which requires substantial efficiency and making

trade-offs between sensitivity and specificity• e. g. categories of genes vs. “gene agnostic” approach?• Especially important given time & resource intensive nature of analysis

• Special considerations in different populations?– e.g. minorities and traditionally underserved groups

• From data analysis to differing values regarding genomic information


Discovery (e.g. novel genes for phenotypes) is “incidental” to the main goals of CSER

Scope of CSER at 9 Sites

• Clinicians Involved: 200• Non-Health Care Professionals: 177• Number of Institutions 25• Peer reviewed publications 31• Non-peer reviewed publications 8• Posters at regional/Ntl/Interntl venues 5• Presentations 35

Patient Focus of Individual CSER Sites• Baylor

– WES of tumor and germline in childhood cancers• Brigham & Women / Harvard

– Randomized trial of WGS in healthy adults & those with suspected hereditary cardiomyopathy• Children’s Hospital of Philadelphia

– WES and WGS in four cohorts of children with pediatric disease• SNHL, ID, mitochondrial, sudden cardiac death

• Dana Farber / Broad Institute– WES of tumor and germline in assorted cancers

• University of North Carolina at Chapel Hill– Randomized trial of WES in adults and children with wide variety of likely genetic diseases

• Cardiac disease, neurodevelopmental disorders, dymorphology, cancer, ophthalmology diseases

• University of Washington– Randomized trial of tumor and germline WES to compare usual care vs. provision of genomic information in adults

with CRC

• Hudson Alpha (new)– Children with intellectual dysfunction

• Kaiser (new)– Adult couples in context of preconception carrier screening

• University of Michigan (new)– WES of tumors in adults and children with advanced cancer

Somatic tumor sequencingGermline only analysis

Aggregate CSER DataReported 9/12/2013

• Total Enrollment: 657– Adults: 429

• 256 female / 173 male– Children: 228

• 107 female / 121 male

• By race– Adults/Children (%)

• American Indian/Alaska Native 7/6 (1.6%/2%)• Asian 10/5 (2.3%/2%)• Native Hawaiian other Pacific Islander 0/0 (0%/0%)• Black or AA, non-Hispanic 19/20 (4.4%/8%)• Hispanic 7/45 (1.6%/19%)• White, non-Hispanic 294/111 (69%/48%)• Not reported or other 94/51 (22%/22%)

Aggregate Data of # Sequenced & Results Returned

• Total Sequenced: 315– Adults 199– Children 116

• Number of participants “with variants /genomic findings”*– Adults 127 (63%)– Children 80 (69%)

• Results returned** in diagnostic context– Adults 28 (14%)– Children 59 (51%)

• Results returned*** as “incidental findings”– Adults 20 (10%)– Children 5 (4%)

•*High/ Variable meaning

•**Low due to those in progress

•***Incomplete consensus on criteria for return

Very rough figures, aggregating germline & somatic analysis

Early tumor report data

Category SubjectsI 2/58 (3%)II 14/58 (24%)III 22/58 (38%)IV 57/58 (98%)I or II 16/58 (28%)I, II or III 32/58 (55%)only IV 26/58 (45%)

Proven or potential clinical utility

Any “cancer gene”mutation


NUMBER OF SUBJECTS with mutations in each category

First 21 CanSeq Patients

• Somatic Alterations: – Curated 82 alterations in 32 genes– Half of these are in 3 genes (KRAS, APC, TP53)– Voted to return 68 out of 82 (83%)

• Germline Alterations:– Curated 88 alterations in 50 genes– Nearly all alterations are unique– Most are VUS– Voted to return 11 of 88 (12%)

NCGENES First 94 cases

• By clinical category– Neuro = 43 cases– Cancer = 14 cases– Cardiology = 8 cases– Ophtho = 29 cases

6 positive (~14%)12 uncertain (~28%)25 negative (~58%)



10 positive (~43%)2 VUS (~9%)11 negative (~48%)





Case of Interest IChallenges of IFs

New mutation(s)?

Chance?

Case of Interest II

• Subject dxd at age 6 with “hereditary spastic paraplegia”– Confined to crutches and wheelchair– Painful spasticity on daily basis

• At 36 no alternative dx had been found in spite of multiple referrals and evaluations nationally


• Rx with oral L-dopa often effective – Dramatic response

Occasional Therapeutic Implications

Case of Interest III






Assorted Challenges Highlighted by CSER Sites in Last Update

• Timely analysis– Especially in setting of diseases with poor prognosis (Baylor, Dana Farber)– When Rx/EOL decisions hinge on analysis (Baylor, Dana Farber)

• Disappointment among some participants when not randomized to receive maximal results (BW)

• Construction of user-friendly reports for full range of users (providers to patients)• Curation and signing out variants

– Complex and time consuming (everyone)– Inaccurate public databases of variants (everyone)– Analytic approach (gene lists vs. gene-agnostic approach (CHOP and UNC)

• Difficulty of providing informed consent given inherent clinical time constraints• Difficulty of / need for independent CLIA confirmation• Eventual expansion of phenotypes associated with mutations• Proliferation of available commercial tests of large panels – how to navigate? • Insufficient knowledge about penetrance of mutations, especially in context of

non-target information

Maintaining a Clinical Focus• MPS is (just another) highly complex medical test• Claims that “soon everyone will have their whole genome

sequenced” are typically predicated only upon its low cost– Even if “free”, perceived low cost is an illusion

• Indiscriminate application of medical tests is very expensive– Morbidity/mortality to individuals (e.g. PSA screening)– Expense to individuals & society

• It should be applied as are other medical tests:– When and if the situation warrants– With targeted genomic analysis in most settings

• Targeting could be actual (e.g. capture) or virtual (at the analytic stage)

– WGS/WES useful in selected diagnostic settings and in research

• Think before burdening the system with a flood of extraneous information that begs for misinterpretation

• CSER is generating data to guide its optimal use

Thank You

[email protected]

Venimus, Vidimus, Secuti Sumus

CSER Plenary Session Thursday, 10/1013 from 10:30-11:30

Hey Lucia and Brad,I’m starting to think about the plenary talk that I’m giving on Thursday, 10/10, from 10:30 to 11:30.Would you look at the rough outline and see what you think? I want to make sure that I talk about things that will be of interest to everyone.I will need help from you guys getting the aggregate data. I figure since the quarterly reports are due 9/16 I might be able to get you to help me soon after that with aggregation. I’m not a procrastinator and like getting stuff done in advance!Finally, what other things do you think I should talk about? Please feel free to make suggestions; I’d appreciate your input.•An overview of the CSER goals/rationale. I don’t think I’ll provide many details on each group’s individual project, but will likely have a slide or two that summarize the overall approach and patient mix for each site.•Total numbers enrolled•Total numbers analyzed•Demographics including ethnicities, age, sex•Phenotypic/clinical categories of subjects being examined•Diagnostic hit-rates broken down by clinical/phenotypic category

With some degree of granularity with regard to results that are clearly positive vs. possible explanations vs. VUS vs. negative resultsI’ll probably take a slight diversion and briefly discuss different approaches to analysis, which will help me keep the interest of the informatics folks

•Frequency of incidental findings deemed reportableI’m assuming that each site reports this information?

•A bit of the ELSI experience thus far, recognizing that it will be largely anecdotal at this early stage•I’ll likely discuss a few individual cases that illustrate broader principles

e.g. a finding we have of a reportable incidental finding and the interesting family history data obtained upon reporting that back to the subject)our case of the highly therapeutic finding of dopa-responsive dystonia dxother illustrative cases that other groups might have

Thanks!

> Sounds great. How about if we leave the precise time (20 vs 15 minutes) a bit flexible for now and as we start to craft our talks we can refine if needed? For now I'll plan on ~20 minutes. I won't have some of the aggregate data for a couple of weeks so things are still up in the air a bit.>

Based on our current data (n=70 tumor exomes), we would estimate that we detect a tumor exome finding that could potentially change pharmacologic treatment of the patient sequenced (with FDA approved drugs or as part of a clinical trial) in the event of tumor recurrence in 25% of pediatric solid tumor cases. The context/caveat to this estimate is that we are sequencing newly-diagnosed pediatric solid tumor patients, who are generally treated initially on or following relatively well-defined protocols, such that the tumor exome data are generally received while the patient is completing their initial standard treatment. Thus in our project the utility of the tumor exome data is being evaluated in the clinical context of recurrent or refractory tumors.Please see attached for several summary slides in case they would be of any use at the meeting.BestWill

Hey,The gene is BBS9, and her number is 141 if you want to look her up. She has a clinical diagnosis of non-syndromic cone-rod dystrophy. She had two variants that were not reported in the literature (but predicted damaging by Polyphen :-) The two rare variants are BBS9 c.1255C>G (p.Pro419Ala) and BBS9 c.2153A>T (p.Glu748Val). The BBS9 gene has not been associated with cone-rod dystrophy (CRD), but of course, CRD is a feature of BBS. Additionally, there are other BBS genes that are associated with non-syndromic retinitis pigmentosa, so we believe her CRD is associated with the two variants identified in her BBS9 genes. We don't have info on phase of the two variants, and we probably won't get the opportunity to study family members.Let me know what additional information you would like on her!

Following slides are from ACMG talk on

lessons learned

Evidence is Needed to Guide Use of New Medical Modalities

• The advent of MPS quickly transformed genetic and medical research

• However, the transition to clinical implementation is complex

• Ultimately, evidence-based guidelines need to be formulated for use

Data Needs for Successful General Implementation

• What are the patient characteristics that signal potential utility (or lack thereof) for applying genome-scale germline sequencing?– Diagnostic “hit” rates in different clinical populations

• What are the best approaches to analysis of data?– Striking the right balance between sensitivity and specificity

• Categories of genes vs. “gene agnostic” approach?• Especially important given time & resource intensive nature of analysis

• Special considerations in different populations?– e.g. minorities and traditionally underserved groups

• From data analysis to differing values regarding genomic information


• How to adjudicate clinical meaning of variants?– ClinGen Project

People Want (and Don’t Want) Off-Target Info for Predictable Reasons

• Most important reasons for wanting to learn IF– To get information about family’s possible health risks– Motivation to do things to prevent or delay future health

problems– To contribute to research

• Most important reasons for NOT wanting to learn IF– Too upsetting to find out about risk for a serious health problem– Information would not give definite answers– No way to cure or prevent health problems

Real Data are Better Than Hypothetical DataPreliminary Findings

• 15/18 indicated they ‘probably’ or ‘definitely’ will ask for at least some IFs– Pharmacogenomics being the most “popular” category

• Our study design asks that subjects make a phone call to obtain off-target information– 6/18 have requested findings to date

Case of Interest I






• 36 yo dxd at 6 with “hereditary spastic paraplegia”– Confined to crutches and

wheelchair for decades– Painful episodes of

spasticity on daily basis


• Dramatic response– Walking without

crutches, free of painful daily symptoms

Case of Interest IIOccasional Therapeutic Implications

Tentative & Preliminary Conclusions

• Diagnostic indication will be critical to determining the utility of WES in the clinic– Wide range of yields for varying conditions

• Retinal disorders vs. Cancer– “distributed” heterogeneity vs. “asymmetric” heterogeneity

• WES vs. pre-defined panels in the clinic?– Panels may be actual or virtual– In knowledge-rich domains and when gene lists are

relatively small (e.g. cancer), panels may offer advantages– Knowledge-deficient domains (e.g. autism, CNS disorders,

dysmorphology and when lists are large, WES is likely to remain advantageous





Case of Interest IIIChallenges of IFs

New mutation(s)?

Chance?


• Inflicting medical care on the healthy requires great caution

• Healthy people arguably have less to gain than sick people • Different relationship between provider & recipient

– The individual isn't typically seeking us out • Benefits are less obvious

– “Great news! You didn't get sick!”• The downsides are easy to see

– All interventions have downsides• The prior probability that any given genomic finding in an

individual is clinically important is low

Genomic Screening of the General Population

Aggregate Data of # Sequenced & Results Returned

• Total Sequenced: 315– Adults 199– Children 116

• Number of participants “with variants /genomic findings”*– Adults 127 (63%)– Children 80 (69%)

• Results returned** in diagnostic context– Adults 28 (14%)– Children 59 (51%)

• Results returned*** as “incidental findings”– Adults 20 (10%)– Children 5 (4%)

•*High/ Variable meaning

•**Low due to those in progress

•***Incomplete consensus on criteria for return

Very rough figures, aggregating germline & somatic analysis

First 21 CanSeq Patients

• Somatic Alterations: – Curated 82 alterations in 32 genes– Half of these are in 3 genes (KRAS, APC, TP53)– Voted to return 68 out of 82 (83%)

• Germline Alterations:– Curated 88 alterations in 50 genes– Nearly all alterations are unique– Most are VUS– Voted to return 11 of 88 (12%)

cser clinical sequencing exploratory research

Documents

clinical carebasic science

gainsaidfor medical

different endeavors

heterogeneous values

clinical practicethe

foundation of medical

premature translation

high misapplication