19.02.19 Interpretation of unclassified variants Flashcards
Critical gene information for variant interpretation?
- Patients phenotype/gene disease link
- Mode of inheritance of gene
- Mutational mechanism
- Protein structure/function
- Strength of gene-disease relationship
Important variant information to gather for variant interpretation?
Follow ACMG guidelines
1) population data - absence from gnomAD/allele freq too high for disease
2) computational and predictive data - prediction is pathogenic or benign, missense where only truncating variants cause disease, synonymous, effect on splicing, variant that changes protein length, inframe indels
3) Functional data - support pathogenicity or being benign
4) Segregation data - no segregation with disease, or segregates with disease in multiple families
5) Inheritance - de novo
6) Allelic data - in trans with a pathogenic variant, bi-allelically inherited
7) Other databases - ClinVar, disease specific ones
8) Other data - previously found in another case with an alternative cause, patients phenotype highly specific, variant discussed at MDT
When are MDTs useful?
- When doing exome or whole genomes sequencing
- Purpose is to assess whether patients phenotype ifts and to ascertain the variants contribution to the clinical presentation
ACMG five class system
1) Pathogenic - >99% probability of being disease-causing
2) Likely pathogenic - >90% probability of being disease causing
3) VUS
4) Likely benign - <10% probability of being disease causing
5) Benign - <0.1% probability of being disease causing
- Class 4/5 are clinically actionable
- Class 3 spans a very wide range (could be nearly class 2 or 4 depending on one extra bit of info)
Analysis steps - Literature search
- Has it been reported before?
- What was the associated phenotype?
- How did they classify it?
Analysis steps - Mutation databases
- In house database (Gemini freq, in alamut, shire)
- HGMD - contains disease causing mutations
- Locus specific databases - BRCA share for BRCA1/2, InSIGHT for colon cancer, RettBASE, LOVD
- Decipher - CNV and SNV data from over 18,000 cases, mostly related to DDD project
Analysis steps - Population databases
- GnomAD - exome and whole genome sequences of unrelated individuals
- Individuals with severe paediatric disease have been excluded so is used as a useful reference set
- Caution - need to consider ethnicity of patient and the fact that laster onset, variable phenotype/penentrance variants may be present in the dataset
Analysis steps - Co-occurrence with known deleterious mutation
- If you find a UV and a known pathogenic variant in trans in a patient with a dominant disorder, that is evidence the UV is not pathogenic (i.e. where a second variant would be lethal)
- Critical to establish phase
- Be careful to look closely at phenotype
Analysis steps - Co-segregation with the disease in the family
- Perform segregation studies to see if VUS segregates with disease
- Good for ruling out pathogenicity
- Limitations due to phenocopies (similar phenotypes in families but with different aetiology), and partial penetrance
- Family structure is also important - issues with non-paternity, not enough family members to test, living in different regions etc
- Be cautious that a variant that looks like it segregates with disease may not and may just be in linkage disequilibrium with it
- Non segregation is considered strong benign evidence
- Co-segregation with disease in multiple family members can be supp, mod or strong evidence
- Use Jarvik and Browning et al, 2016 to calculate strength of evidence that you have (its the number of meioses that you have, not the number of family members)
- Can also use multiple families to increase strength of evidence further
Analysis steps - De novo
- For AD diseases, inheritance info can provide strong evidence
- Caution regarding incomplete penetrance, mosaicism, non-paternity, and how specific the phenotype is
Analysis steps - In silico predictions
1) Species conservation
- disease causing mutations are more likely to occur at positions that are conserved throughout evolution
- Because these variants are under strong purifying selection and eliminated from populations - therefore suggesting that amino acid is critical for proper gene function
- More variable regions may show areas that are under less severe selection constraints and therefore where changes may be more tolerated by natural selection (therefore you can change the amino acid without negatively impacting on protein function)
- ACGS recommend that you use an alignment that includes the full length of 8 orthologous genes and at least 5 mammalian orthologues plus chicken, frog and fish
2) Protein based predictions
- Used to predict effect on protein
- Most common are SIFT, PolyPhen and Align GVGD
- Align GVGD - combines Grantham matrix with sequence alignments
- SIFT and PolyPhen are based on multiple sequence alignments
- PolyPhen also takes into account 3D structure
- In future, for missense variants, a meta-predictor tool (e.g REVEL, GAVIN) will replace the use of multiple individuals ones
3) Splicing
- Intronic, synonymous or non-synonymous variants can have splicing effects through changes on splice donor and acceptor sites as well as splicing enhancers and silencers
- GeneSplicer, NNSplice etc can predict this
- Testing for an affect on splicing should be considered especially when AG and GT dinucleotdie sequences ar formed
Analysis steps - RNA studies
- Used to investigate the potential effect of change on normal splicing by studying cDNA generated from a fresh mRNA sample. Need to consider…..
1) Normal isoforms - what is normally expressed
2) Expression of mRNA on blood - is it? or may need another tissue to test
3) Quality of RNA - it degrades very quickly
4) NMD - if splicing change creates a premature termination codon the mRNA may be subject to NMD (i.e. variant splicing product would not be present, or at low levels)
Analysis steps - LoH
- Important for tumour supressor genes, where LoH can provide evidence for or against pathogenicity
- If VUS found in germline DNA, LoH of the normal allele in tumour tissue suggests pathogenicity
- Things to consider
1) If the variant is not seen in tumour, suggests its benign
2) Even if LoH is seen, may not be a causal variant
3) presence of normal tissue in tumour may obscure results
Analysis steps - Functional studies
- Reliable functional assay is one of the best ways to confirm pathogenicity
- However very hard to get good functional data - be careful to check diagnostic evaluation of an assay before using it as evidence
- Can have in vitro or in vivo assays, and require you to measure something that is associated with function
- If functional assay is negative - was it performed correctly? with the right controls? measuring the right thing?
Analysis steps - Enzyme analysis and immunohistochemistry (IHC)
- For some disorders can run disease-specific enzyme tests which may provide evidence (e.g. for FAH)
- IHC staining can show presence or loss of a protein in tumour tissue which could help classification
