1. Re-annotation of Genomic Sequence
Why taking another look is important
CGM Annual Symposium
Secondary Findings from genome wide testing
November 7th 2016
Christian Marshall
Genome Diagnostics
Centre for Genetic Medicine
The Centre for Applied Genomics
The Hospital for Sick Children

2. Hypothetical case: 5 year old male with unexplained Intellectual Disability and some distinct craniofacial features. Previous microarray testing negative.
Genome interpretation is a snapshot relying on current evidence and phenotype
How are variants interpreted? When is a secondary finding actually related to the phenotype and when is it predictive? When do we reanalyze?
Evidence and phenotype
WES ordered and no variants explaining the primary phenotype detected
A rare missense variant in BRCA1 is found with some evidence of cancer susceptibility, reported as a secondary finding.
2011
Ordered re-analysis of the WES. Pathogenic change in PACS1 reported
New disease gene causing Schuurs-Hoeijmakers syndrome that matches the primary phenotype.
2013
Audiological testing revealed bilateral high-frequency sensorineural hearing loss (SNHL). He also had an episode of blood in his urine in the past.
His mother, who is 40 years old has developed hypertension, and has been found to have microhematuria and proteinuria.
A rare COL4A5 VUS is identified upon testing. Subsequent renal biopsy on the mother reveals pathology consistent with a diagnosis of XL Alport.
BRCA1 variant detected as a secondary finding but accumulating evidence has shown it is likely benign.
2016
Rebekah Jobling

3. Strategies for Diagnostic Genetic Testing using Next Generation Sequencing
Patient with heterogeneous disorder
Targeted Gene panel
Gene Panel
High coverage
Optimized
Easy to analyze and interpret
Low complexity but limited flexibility
In Silico Targeted Gene panel + non-coding
Whole Genome
No gene bias
Easy to prioritize known genes then expand search
Non-coding and Structural variants available
Find medically actionable variants and allow gene discovery
High Complexity
Actionable variants
Gene Discovery
Structural Variants
Non-Coding Variants
In Silico Targeted Gene panel
Whole Exome
Little No gene bias
Easy to prioritize known genes then expand search
Find medically actionable variants and allow gene discovery
High Complexity
Actionable variants
Gene Discovery
Trend is towards more ‘genomic’ sequencing, a more hypothesis free approach for clinically heterogenous disorders.

4. Increasing diagnostic yield in heterogeneous disorders
52%
44%
25%
20%
16%
10%
11% 5% 3% 0%
Neveling et al. Hum Mut. (2013)

5. Increasing diagnostic yield in heterogeneous disorders
In a clinically heterogeneous paediatric cohort referred for microarray analysis:
Microarray diagnostic rate was 8%
All targeted tests (average of 3 per patient) diagnostic rate was 13%
WGS (CNVs and SNVs) diagnostic rate was 34%

6. Overview of Next Generation Sequencing
Sequence Read Generation
Sequence Read Alignment
Base calls
Aligned reads
Reads
ATGAGCTAC
Reference Genome
Variant Annotation
Variant (SNV + CNV Calling)
ExaC
1000G
PolyPhen
SIFT
OMIM
HGMD
g.chr15:44,955,876C>T
Gene
Each step can effect final variant list and interpretation
Although there are technology and algorithm improvements, updated annotation is the most likely to influence variant interpretation

7. Variant call format (vcf)
Callers will output data in vcf – variant call format which the standard for NGS experiments
Consists of a header and a data section - now what?

8. Variant Annotation – add context to the variants
Effect of variant on gene
Prediction of variant effect (amino acid changes or splicing)
Population Frequencies
Potential Disease association

9. AnnotatedVariant calls (.vcf)
Variant Annotation
Population Freqs
Transcript conseq.
Pathogenicity Pred
Disease DBs
dbSNP
Annovar
PolyPhen/SIFT
OMIM
AnnotatedVariant calls (.vcf) …
Variant calls (.vcf) …
1000G
SNPeff
PhyloP
HGMD
ExAC
Ensembl VEP
Splicing predictors
ClinVar
Last part of the analysis pipeline -> needed for variant interpretation
Requires constant error checking and updating as databases are updated
Annotation is customizable
Standard and case independent

10. Individualized Annotation
Clinical Notes:
Male 11 year old patient with lipodystrophy, developmental delay (speech delay). Also has developed umbilical hernia, anxiety disorder and renal microcysts and hyperkalemia.
Human Phenotype Ontology:
HP: : Lipodystrophy
HP: : Umbilical hernia
HP: : Renal cyst
HP: : Global developmental delay
HP: : Hyperkalemia
Standardize clinical descriptions and use to prioritize variants in an iterative way with phenotypic information
Note: OMIM and phenotype can be very noisy – in our experience using the HPO in an automated way may NOT give you the causative variant

11. Variant Interpretation in Genomic sequencing
Variants
Phenotype

12. Variant Interpretation Guidelines
Describes updated standards and guidelines for the classification of sequence variants using criteria informed by expert opinion and empirical data
Systematic framework to classify variants with standard terminology
5-tier system using different ‘bins’ of evidence
ACMG. GIM. (2015)

13. Evidence for Pathogenicity
Scoring system with varying degrees of evidence
Optimized for Rare Mendelian Disorders
ACMG. GIM. (2015)

14. Scoring Variants
Pathogenic
1 Very Strong AND - 1 Strong OR
- ≥2 Moderate OR
- 1 Moderate and 1 Supporting OR
- ≥2 Supporting
2 Strong
1 Strong AND - ≥3 Moderate OR - ≥2 Moderate and ≥2 Supporting OR - ≥1 Moderate and ≥4 Supporting
Likely Pathogenic
1 Very Strong or 1 Strong AND - ≥1 Moderate OR - ≥2 Supporting
≥3 Moderate
2 Moderate AND ≥2 Supporting
≥1 Moderate AND ≥4 Supporting
Benign
1 Stand-Alone
≥2 Strong
 Uncertain Significance
Everything else!
Likely Benign
1 Strong and 1 Supporting OR
≥2 Supporting
Known Loss of function variant: PVS1, PS4, PM2 = Pathogenic
Variant >5% in the general population: BS1 = Benign
Extremely rare missense variant with no other evidence: PM2 = VUS

15. Variant of Uncertain Significance
Variant Analysis and Interpretation
Reported primary
Reported secondary
Not disease causing
Disease causing
Benign
(polymorphism)
Likely Benign
Variant of Uncertain Significance
Likely Pathogenic
Pathogenic
Some Considerations:
Evaluation of Secondary Findings
Required required evidence to call a variant pathogenic should be higher in the absence of phenotype
Genes of uncertain significance (GUS)
If variant is found in a gene with no known associated phenotype, many of the classifications do not apply -> variant of uncertain significance
Reanalysis
Laboratories encouraged to develop approaches to give patients and providers more efficient access to updated information
At the very least encourage periodic enquiries from heath care providers
Discordant calls: although a good framework, many interpretations are discordant across laboratories!

16. Why is it important to go back?
Population Freqs
Transcript conseq.
Pathogenicity Pred
Disease DBs
dbSNP
Annovar
PolyPhen/SIFT
OMIM
AnnotatedVariant calls (.vcf) …
Variant calls (.vcf) …
1000G
SNPeff
PhyloP
HGMD
ExAC
Ensembl VEP
Splicing predictors
ClinVar
Many different (external) databases with constant updating
Can significantly influence variant (and gene) reporting

17. Disease Gene Annotation - HGMD
HGMD genes per year
HGMD variants per year
7674
190K
The ability to generate vast amounts of data have lead to more disease genes and more variants that are pathogenic

18. Disease Gene Annotation - Clinvar
Community based: clinical significance is reported directly from submitters.
Currently holds > submitted variants with interpretations for >
~4800 genes are represented

19. Disease Gene Annotation - OMIM
Approximately 250 gene–disease and 9,200 variant– disease associations are reported annually
Necessitates re-annoation and reassessment
Wenger et al. GIM. (2016)

20. Re-annotation in clinical WES
Re-evaluation of non-diagnostic exomes for 40 cases
A definitive diagnosis was identified for 4 of 40 participants (10%) upon reanalysis -> mainly due to new gene discoveries
No mention of secondary findings
Wenger et al. GIM. (2016)

21. When do we reanalyze genomic data?
Certainly a benefit to re-annotation and reanalysis of genomic data
Laboratories: rely on providers for phenotypic data and for updates to this information to be considered at reanalysis.
Providers: rely on laboratories to report relevant variants and to update variant interpretation based on new information
General Policies for reassessment are laboratory specific:
Some labs do regular variant assessment automatically and recontact provider
Majority of labs will reassess variants upon request, reanalysis of WES (may be a fee)
Policies are geared towards primary findings, reassessment of reported secondary variants is possible but no standard in the field for systematic reanalysis of genomic sequence for secondary findings

22. Variant Interpretation changing with time
Rare Variants with established pathogenicity
Pathogenic
Very (ultra) rare variant need family and phenotype + functional work to interpret
“Susceptibility”? Slight increase in cases? Some functional evidence
Likely pathogenic
Report 2o
Variant of Uncertain Significance
Rare variant, same frequency in cases and controls
Rare variant, same frequency in population as data accumulates
Variants >5%
Report 1o
Likely benign
Benign
Time
Variants can have a different trajectory as evidence accumulates
Large numbers are imperative but also important to have phenotype information and thus the interaction between laboratories and physicians is crucial

23. Summary and Observations
Genetic investigations for diagnostics trending towards genomic rather than targeted testing due to increased diagnostic yield
Variant Assessment and categorization follows a systematic scoring system based on evidence at the time of analysis
Required required evidence to call a variant pathogenic is higher in the absence of phenotype
Genome Annotation is based on evolving databases and thus variant interpretation also evolves
Laboratories need to develop an efficient process for updating annotation and reassessing variants
In genomics era of testing, the collaboration and communication between the laboratory and physicians is crucial; assessment needs to be iterative in the context of flow of detailed phenotype and variant evidence

24. Acknowledgements
SickKids/Centre for Genetic Medicine, Genome Clinic Working Group
SickKids Genome Diagnostics
James Stavropoulos
Peter Ray
Raveen Basran
Lianna Kyriakopoulou
Rebekah Jobling
SickKids Centre for Applied Genomics
Stephen Scherer
Daniele Merico
Centre for Genetic Medicine
Ronald Cohn
Sarah Bowdin
Steve Meyn
Nasim Monferad
Cheryl Shuman
Robin Hayeems
Chris Carew
Adel Gilbert