The Genetics Clinic of the Future a Dutch consortium ......The Genetics Clinic of the Future –a Dutch consortium approach to diagnostic Whole Genome Sequencing Ies Nijman, University

The Genetics Clinic of the Future – a

Dutch consortium approach to

diagnostic Whole Genome Sequencing

Ies Nijman, University Medical Center Utrecht, The Netherlands

University Medical Center Utrecht

• One of the 8 Academic hospitals in the

Netherlands.

• >1000 beds

• >30.000 patients per annum.

• >11.000 employees

• Merge of Academic hospital, Children's

Hospital and university Medical faculty.

• JCI quality accredited since 2013

• UMCU Department of

Genetics

• 10.000 patients annually

• 340 professionals

• NGS facility, including

data analysis and

interpretation

• Reimbursement

agreements (DBCs)

• ISO 15189 certified

Traditional clinical genetic care

SymptomsClinical

diagnosisDifferential diagnosis

Molecular diagnosis

rubenmaalman.nl/genetics

Australian study on rare disease diagnostics

• Survey (1014 completed responses) on patients with rare disease (freq <1:2000 of which 80% is estimated to be genetic)

• ~50% received ‘at least one incorrect diagnosis’.

0

5

10

15

20

25

30

35

40

1–2 3–5 6–10 11 or more

% o

f p

atie

nts

# of doctors

Number of doctors seen to get a confirmed diagnosis (n = 735)

0

5

10

15

20

25

30

<3 months 3–12 months 1–5 years 5–10 years >20 years

% o

f p

atie

nts

time to diagnosis

Time to diagnosis (n = 718)

Changes in clinical genetic care

GCOF – the elements

Pilots

• Focus Clinic for Children with ID

• Genetics First for Primary Immune Deficiencies

• CardioGenetics

• NICU: neonatal intensive care. High priority exome+array diagnostics with 11 days t-a-t.

GF-PID

Consortium

GCOF: Children with ID.

Now

Total diagnostics

Genetic diagnosis

(~6-8 weeks)*

Total diagnostics

DiagnosePediatrician

Child Neurologist

Pediatrician

Child NeurologistDiagnose

Clinical Geneticist

Clinical Geneticist

Future

Genetic diagnosis

(~6-8 months)

HTA: health technology assessment

• 2 PhD projects started monitoring patient during health care process:

– Clinical genetics: how effective is genetics first?

– Oncology: hoe effective is genetics guided targeted treatment?

• Involvement of insurance companies.

Impact WES on cost and medical decision making

Before WES € 12.26 € 10.47 € 9.98

After WES € 2.62 € 1.79 € 2.06

YES NO VUS

• Cost study of 370 patients with intellectual disability receiving WES• All health care consumption before and after WES collected

• Around 40% of the average costs before WES were geared towards obtaining a diagnosis, i.e. interventions in the categories Diagnostics, Lab, Consultations and Genetics (“yes”: €7,621; “no”: €6,203; “uncertain”: €7,490)

• Cost per day after WES are significantly decreased in every diagnosis category

Conclusion• WES first results in cost savings• WES has a distinct “end of trajectory” effect

Cost effectiveness

• Sub Project: the genome first approach

– Basic principle of GCOF: the patients’ DNA data

is available very early in the process as

fundamental ‘biomarker’.

– This DNA data needs to as broad as possible:

whole genome sequence (WGS) - one test fits all.

• Replace WES, targeted sequencing, array CNV

analyses, karyotyping, MLPA

Sub-national scaling of diagnostic sequencing

• Dutch Society Clinical Genetics Laboratorydiagnostics coordinated collaboration between Academic hospitals of Amsterdam (3), Groningen, and Utrecht.

• Replace panel/exome sequencing by whole genome sequencing.

• Outsource sequencing to large scale illumina Xten facility

• Consolidate local analyses into single bioinformatic pipeline.

• Validate procedure and pipeline following quality certifications.

• Variant Annotation and interpretation done by hospital Systems already in place (Cartagenia)

• Implement WGS flow in local hospitals.

• Scale up to cover 60% of NGS based analyses (11.000 pts/year) in 2019.

Validation results SNPs & InDels

• Validated on Genome-in-a-bottle sample NA12878 and ‘truth’ set NIST 3.2.2

• 10 runs show:

– Robust and stable results with high sensitivity and precision on ‘standard’ 30x coverage genomes.

– Results are already an improvement over exome sequencing performance, while filter setting can still be tuned to decrease false negatives.

– Doubling coverage to 75x does not significantly increase sensitivity or precision.

GENOME EXOME target

All variants Filtered variants All variants Filtered variants

SNP Precision 99.36% 99.79% 99.02% 99.63%

Indel Precision 96.60% 97.08% 96.86% 97.26%

SNP Sensitivity 99.73% 98.55% 99.69% 99.04%

Indel Sensitivity 96.63% 96.41% 97.50% 97.25%

Coverage analysis

• In a clinical/diagnostic setting negative results are equally important.

• Absence of proof is not proof of absence!

• tripling coverage expands the ‘diagnosable region’ only slightly (1-2% @15x)

20

30

40

50

60

70

80

90

100

20 40 60 80 100 120

frac

tio

n o

f p

ane

l co

vere

d

mean genome coverage

fraction of target panels covered

1x 10x 15x 20x 30x

35-40x coverage is the sweetspot for a broad

diagnosable area

• To get a target covered at 15x for >99%: 35-40x genome coverage is the sweetspot.

• Adding more coverage does not effectively fill the missing regions nor does it significantly increase sensitivity or precision

• Better quantifier for data usability under investigation: Genotype Quality

Replacing array based CNV by WGS

• WGS allows detection of both balanced as well as unbalanced CNV events at potentially higher resolutions

• Parallel diagnostic tracks of WGS and array is complicated, time consuming and expensive.

WGS

Array

title

• Blaat

Validation results of WGS based CNV diagnostics

• Validate array diagnostics vs Whole Genome Seq

• 5 tools tested (CNVnator, qDNAseq, lumpy, Delly, Manta)

• 24 samples tested:– 3 samples with balanced events (inv, 2x translocation)

– In 18 samples 36 class3-5 (VOUS, LP or P unbalanced events)

– In 21 samples 153 class 1-2 (benign-likely benign unbalanced events)

• qDNAseq best performer, followed by CNVnator.– qDNAseq calling on ‘callable regions’ of the WGS.

– qDNAseq: 92% sensitivity class 3-5 (45% class 1-2).

– CNVnator: 83% sensitivity class 3-5 (57% class 1-2)

Working with CNV data

• Array diagnostics based on reference/population data: need for NGS based reference database of calls to effectively filter false positive/benign events.

• Adjustment of analysis software

• Interpretation of genotypes combining CNVs and SNP/Indels.

Data sharing is a must

• The abundant local knowledge and expertise on data needs to be shared within the professional community!

• Sharing of variant interpretations and forming a consensus allows faster and better diagnostics

• Sharing allele/GT frequencies across labs next to using public databases will reduce the amount of new variants waiting for interpretation.

Managing the costs..

• WGS on current (Xten) platform still expensive for diagnostics

• Upcoming platforms might bring sequencing costs down…

• …But informatics costs will not change.

• Soon, informatics costs will be higher than sequencing costs..

Conclusions

• WGS can be successfully implemented in routine diagnostic practice and replace WES and array CNV.

– Requires (large) enterprise informatics next to lab automation and bioinformatics solutions.

– Full potential not unlocked yet: primary focus on continuation of current tests

• Next: high resolution CNV (indels 100-5000 bp), Balanced SVs

• Next: Non-coding, regulatory and splice variants.

• Layered analyses allow answering multiple diagnostic questions faster while limiting incidental findings.

• Data/knowledge sharing essential for efficient and consistent diagnostics.

• We need machine readable phenotyping integrated in hospital systems and data analysis

• Investigations on use and impact of WGS on (medical) society are still required

• Regulations need to be adapted to the whole genome scenarios.

Acknowledgements

Daphne van BeekMatt HestandDaoud SieMarja Van StralenJanneke WeissEric Sistermans

Ies NijmanRobert ErnstPatrick van ZonHeleen Schuring BlomTerry VrijenhoekHans Kristiaan Ploos van Amstel

Stef van LieshoutTed BradlyEwart de BruijnIes NijmanEdwin Cuppen

Fred Van RuissenFrank Baas

Pieter NeerincxMarloes BenjaminsBirgit SikkemaRichard Sinke

Jelle ten HoeveFrans Hogervorst