the clingen sequence variant interpretation working group: refining criteria for interpreting the...
TRANSCRIPT
The ClinGen Sequence Variant Interpretation (SVI) Working Group-
Refining Criteria for Interpreting the Pathogenicity of Genetic Variants
Marc Greenblatt, Leslie Biesecker, Danielle Azzariti, Jonathan Berg, Sharon Plon, Heidi Rehm, and the ClinGen SVI Working Group
3 June 2016
ClinGen SVI Work Group• Co-Chairs• Les Biesecker (NHGRI) • Marc Greenblatt (U of Vermont)
• Danielle Azzariti (Harvard/Partners)• Jonathan Berg (UNC)• Steven Brenner (UC Berkeley)• Fergus Couch (Mayo Clinic)• Selina S. Dwight (Stanford) • Raj Ghosh (Baylor)• Steven Harrison (Harvard/Partners)• Chris Heinen (UConn) • Alison Homstad (UNC)
• Matt Hurles (Sanger)• Peter Kang (Counsyl)• Rachel Karchin (Johns Hopkins)• Annie Niehaus (NHGRI) • Robert Nussbaum (Invitae) • Sharon Plon (Baylor)• Erin Ramos (NHGRI)• Heidi Rehm (Harvard/Partners)• Tasha Strande (UNC)• Sean Tavtigian (U of Utah)• Kira Wong (NHGRI) • Matt Wright (Stanford)
Case Study of Confusion in Variant Classification
• Clinical hx– Breast CA, 50 yo woman, Ashkenazi Jewish– Mother Breast CA 62, Maternal GM Panc CA 71, Maternal
GM’s sister Breast CA 48
• Genetic Workup– 2008: Mother (-) for the 3 common Ashk J BRCA variants– 2015: Proband, 61 gene panel: “Negative for mutations”– 2016: Mother’s unaffected sister, 25 gene panel (different
testing company): • VUS in ATM• VUS in CHEK2
Case Study of Confusion in Variant Classification
• Reconciling the Genetic Workup– Affected individual “Negative”– Unaffected second degree relative with two VUS’s
• Query the first company regarding the VUS’s– ATM VUS was present, classified as “Benign”, not reported– CHEK2 VUS not present- “but if it had been present, we
would have called it ‘likely pathogenic’”– CHEK2 VUS in ClinVar: Four labs have reported variant, all
say “Pathogenic”
• One family, 2 cancer genes, 2 discordant variants
Background: Need to Standardize Variant Classification
• Conflicting systems among labs, researchers• American College of Genetics and Genomics (ACMG)
panel developed guidelines for evaluating different types of data (Richards et al, Genetics in Medicine, 2015)
• Qualitative evaluation of different data types• Evidence assessed as “Very Strong”, “Strong”,
“Moderate”, “Supportive”• Goal: to develop a system that labs can use, reduce
discordance, instill confidence in evaluating evidence
ACMG Framework for Classifying Variants
Background: ClinGen
• US NIH is funding a coordinated effort to create a Clinically Relevant Variants Resource
• ClinVar: “public archive of reports of the relationships among human variations and phenotypes, with supporting evidence”
• ClinGen: “build an authoritative central resource that defines the clinical relevance of genes and variants for use in precision medicine and research”
ClinGen Sequence Variant Interpretation (SVI) Work Group
• Refine standards for variant interpretation– Assess each data type in the ACMG guidelines– Establish standards for how to integrate data
• Short term goal: Refine, clarify, and modify current ACMG/AMP criteria
• Long term goal: Move to quantitative Bayesian framework
SVI WG Process
• Start with ACMG grid • Standardize current interpretation processes
for each cell in the grid– Sub-groups of 3-5 people address each category – Currently ~20 people involved in the WG– Work with other groups, eg disease-specific WGs
ACMG Framework for Classifying Variants
ACMG Framework for Classifying Variants
First data typesto address:-Population Frequencies
-In silico algorithms
First ACMG Criterion for SVI: “BA1” Using Population Allele Frequency
• “BA1”- Benign, criterion can stand Alone
• Use allele frequency in control population as a diagnostic criterion for “Benign”– Rare that allele frequency >5% is associated with disease
• Harmonize approach of SVI with disease related groups (Cardiomyopathy, RASopathy) so that ClinGen presents a single coherent message
First ACMG Criterion for SVI: “BA1” Using Population Allele Frequency
• Current wording: “Allele frequency is >5% in Exome Sequencing Project, 1000 Genomes Project, or Exome Aggregation Consortium”
• Proposed wording: “Allele frequency is >.05 in any general continental population data set of at least 2,000 alleles for a gene without a gene-specific recommendation.”
Refining Computational Methods
• Algorithms based on sequence, structure• To standardize, issues to address include:
– How to assess sequence alignments, algorithms– How to validate/calibrate outputs
• Criteria for Genes to use to assess algorithms– Variants classified as Pathogenic and as Neutral– Disease phenotype defined clearly– Expert group to assist in classification
Using ClinVar to Identify Genes
• Search ClinVar for genes with large numbers of variants reported
• Apply filters to get missense variants• Use ClinVar “Star” system
– One star- pathogenicity assertion from one source– Two stars- concordant assertions from 2+ sources
ClinVar Genes, N>50 “2-Star Variants”
• Only 5 genes (BRCA1/2, MLH1, MSH2, CFTR) have N>50 missense variants reported in ClinVar with assertions from >1 source
• Exact figures are likely not accurate
• Disease focused groups (InSiGHT, ENIGMA, CFTR2) note more missense variants than ClinVar
• Need to engage disease-specific groups
ClinVar Genes, N>50 “1-Star Variants”Neutral v Path Variants N Genes >20 Neutral >20 Path 10
10-19 Neutral >20 Path 5 <10 Neutral >20 Path 15
>20 Neutral <20 Path 6
Total >50 36
• Disease-specific groups likely would revise numbers (e.g., CFTR includes only 9 Neutral)
• Include genes from many disease categories (cancer, cardio, hearing loss, myopathies, dysmorph, neuro)
SUMMARY
• ACMG framework is a starting point for refining criteria to classify variants
• ClinGen SVI has short and long term goals– Add precision to individual criteria– Create framework for integrating data– Convert to quantitative system
• Work with disease-focused groups
ClinVar Genes, N>50 “2-Star Variants”Gene Total Pathogenic NeutralBRCA1 115 25 90BRCA2 110 9 101MLH1 109 82 27MSH2 60 43 17CFTR 54 52 2
• From Raj Ghosh extracting ClinVar summary data• Exact figures are likely not accurate• E.g., InSiGHT notes >350 Pathogenic, >250 Neutral
missense variants; BRCA2 has >9 Pathogenic• Need to engage disease-specific groups
Short versus Long-Term Goals
• Move from qualitative system toward more quantitative system– Overlay probabilities onto current qualitative
terms (e.g., what is “Strong” versus “Moderate” versus “Supporting”)
– Develop quantitative schema for the future