Interpretation of Genomic Variants**

Question 1

Why is interpretation of genomic variants so important?

Answer 1

It is the responsibility of the Clinical Scientist to provide a clear and unambiguous description of any genomic variants, and whenever possible, an explanation of the clinical implications of the results.

Question 2

Why is interpretation of genomic variants such a challenge and what resources are aviable to assist?

Answer 2

1. Whilst research laboratories have large resources at their disposal to investigate individual variants with advanced computational and functional studies, routine diagnostic service laboratories do not.
2. The interpretation of never-before-seen variants represents a considerable challenge for the diagnostic community.
3. Various professional bodies have published guidelines to assist with variant interpretation.

Question 3

What are the key guidelines for interpretation of SNVs and CNVs?

Answer 3

1. Constitutional Postnatal Chromosomal Microarray Best Practice Guidelines (2011)
2. ACMG Standards and Guidelines for constitutional cytogenomic microarray analysis (2013)
3. Practice Guidelines for the Evaluation of Pathogenicity and the Reporting of Sequence Variants in Clinical Molecular Genetics (2015)
4. ACMG Standards and guidelines for the interpretation of sequence variants (2015)

Question 4

What is the definition of ‘CNV’?

Answer 4

• Redon et al. (2006) defined a CNV as a DNA segment of one kilobase (kb) or larger that is present at a variable copy number in comparison with a reference genome.
• Copy number chages <1kb are called ‘deletions’ or ‘insertions’

Question 5

Why has CNV interpretation become more important in recent years?

Answer 5

• The widespread use of arrays as a frontline test and increasingly NGS technology means these smaller CNVs now being detected and uploaded to CNV databases.
• Consitency in patient care/management relies on standardised a approach to interpretation of CNVs

Question 6

What is meant by ‘CNV classification’?

Answer 6

Distinguishing benign CNVs from disease causing pathogenic CNVs is key to CNV interpretation.

CNVs can be broadly categorised into three classes:

1. Benign copy number changes, likely to be of no clinical significance
2. Copy number variants of uncertain clinical significance (VOUS)
3. Likely pathogenic- Copy number changes involving regions of known clinical significance, or large changes

Question 7

Broadly, what key information is used in the process of CNV classification?

Answer 7

Classification of CNVs is based upon assessment of the region involved, by reference to various internal and external databases and literature

Question 8

How common are benign CNVs in normal population?

Answer 8

• Benign CNVs are found at high frequency in human populations
• Zarrei 2015 - Copy number variation map
  
  • 4.8–9.5% of the genome is copy number variable
  • ~100 genes that can be completely deleted without producing apparent phenotypic consequences
  • Uneven distribution of benign CNVs between chromosomes
• All individuals are typically hemizygous for approximately 30 to 50 deletions larger than 5 kb
• Frequency of CNVs may be greater than 100 per individual (Coughlin 2012)

Question 9

What key database is used to assess CNVs in the ‘normal’ population?

Answer 9

• The Database of Genomic Variants: a curated collection of structural variation in the human genome
• Current version of DGV consists of 55 published studies, comprising >2.5 million entries identified in >22,300 genomes; http://dgv.tcag.ca/dgv/app/home

Question 10

What key criteria are required in order to interpret a CNV as benign? (Miller 2010)

Answer 10

1. Identical CNV inherited from a healthy parent
2. Similar to a CNV in a healthy relative
3. CNV is completely contained within genomic imbalance defined by a high-resolution technology in a CNV database of healthy individuals
4. CNV is gene poor
5. CNV is a duplication (no known dosage sensitive genes)
6. CNV is devoid of known regulatory elements

Question 11

What caveats must the analyst be aware of when interpreting apparently common (benign) CNVs?

Answer 11

• heterozygous deletion of a CNV can be phenotypically benign, while a homozygous deletion can have serious phenotypic consequences.
• heterozygous deletion may also uncover a pathogenic recessive mutation or an imprinted region that is present on the homologous chromosome

Question 12

What key criteria are required in order to interpret a CNV as pathogenic? (Miller 2010)

Answer 12

1. Expanded or altered CNV inherited from a parent
2. Identical CNV inherited from an affected parent
3. Similar to a CNV in an affected relative
4. CNV is de novo
5. CNV overlaps a genomic imbalance designed by a high-resolution technology in a CNV database for affected patients
6. CNV overlaps genomic coordinates for a known genomic imbalance syndrome
7. CNV contains morbid OMIM genes
8. CNV is gene rich
9. CNV is a deletion
10. CNV is a homozygous deletion
11. CNV is an amplification (greater than 1 copy gain)

Question 13

What key database is used to assess CNVs in disease populations?

Answer 13

1. Decipher: Provides size of break points of previously recorded imbalances along with phenotype information and whether the CNV was de novo
2. ECARUCA: Contains cytogenetic and clinical data of rare chromosomal aberration
3. ClinVar: Sequence variants and structural variants with clinical information and clinical significance assigned by the submitter
4. Internal database: importance source of previously seen CNVs and can be used to eliminate platform artefacts

Question 14

When is a CNV classified as VUS?

Answer 14

When there is insufficient evidence to classify as clearly pathogenic or benign.

Question 15

Approx 25-30% of prenatal arrays will identify CNVs which will require follow up studies.

Are CNVs interpreted differently in prenatal testing setting?

Answer 15

The same database tools should be utilised for the interpretation of CNVs however,

The approach taken is to only report de novo CNVs which correspond to a “recognised fully penetrant microdeletion/duplication syndrome or corresponds to a significant chromosomal imbalance”.

Question 16

What is a “Unclassified Variant” (UV)?

Answer 16

A UV is an allele, or variant form of a gene, which has been identified through genetic testing, but whose significance to the function or health of an organism is not known.

Also called variant of uncertain significance (VUS), however this specifically referrs to a UV which has been through the interpretation process and can not be definitively classified as pathogenic or benign.

Question 17

How are UV’s commonly identified and why are they becoming more common?

Answer 17

• UVs are often identified when a whole gene is screened by sequencing as by definition if the mutation is interrogated with a targetted assay then it is unlikely to be a UV.
• Increased numbers of UVs have been identified as next generation sequencing has become the method of choice for many genetic tests and more genes analysed in a more hypothesis-free driven approach.
• The same is true with VOUSs identified by genome-wide CMAs with increasing probe density.
• Moving towards an era of WGS a standardised framework for assessing genomic variants is essential.

Question 18

How do labs typically approach the assessment of a UV?

Answer 18

• Most laboratories have a standard process that is performed when a UV is identified
• Information is collated from many sources into one document and a conclusion is drawn.
• Many laboratories use the software package, Alamut, to accelerate the gathering of information however this is not mandatory.
• It is becoming increasingly important to have clinical input into the UV assessment process via MDT meetings

Question 19

What information sources are interrogated by most laboratories when assessing a UV?

Answer 19

1. Literature search
2. Mutation databases (LSDBs)
3. Population databases (unaffected individuals)
4. Co-occurrence with known deleterious mutation
5. Co-segregation with the disease in the family
6. Inheritance (?De novo variant)
7. In silico predictions
8. Mutation spectrum
9. RNA studies
10. Somatic Loss of heterozygosity (LOH)
11. Functional studies
12. Enzyme analysis
13. IHC

Question 20

How is the gathered information assimilated into a variant classification?

Answer 20

• Best practice guidlines provide a model for classifying variants into groups from 1 (benign) of 5 (pathogenic).
• Historically (ACGS 2015) these have been less prescriptive leaving the final classification down to ‘proessional judgement’.
• More recently the IARC (plon 2008) group ascribed a probability of pathogenicity to required for classification into each group.
• In 2015 the ACMG provided a highly prescriptive set of criteria which must be met in order to classify a variant into each of the 5 groups.

Question 21

What considerations must be taken into account when assessing UVs with regard to literature search?

Answer 21

• It is important to perform a full literature search to see if the UV has been reported before
• Important to ascertain what the author’s interpretation of it is and what studies have been done to come to that conclusion.
• Literature searches should be performed using both nucleotide and protein nomenclature, in addition to using non-HGVS nomenclature (alamut).
• A Google search can also be useful to identify abstracts given at talks or specific-laboratory data.

Question 22

What considerations must be taken into account when assessing UVs with regard to mutation databases?

Answer 22

• Caution is key!
• The quality of databases varies – most informative ones are curated and have relevant comments related to the clinical interpretation of the variant.

Question 23

List some key mutation databases used for assessing UVs (specifically SNVs/Indels).

Answer 23

1. The Human Gene Mutation Database (HGMD®) - links to published cases of affected individuals in the literature also includes functional polymorphisms
2. Locus-specific databases (LSDBs) - There are many LSDBs for many different diseases. Leiden Open Variant Database (LOVD) is a web-portal for accessing many of these.
3. DECIPHER - Interactive database that includes data from >20,000 cases (250 centres) and includes both CNV and SNV data. Includes trio data from the deciphering developmental disorders project (DDD)

Question 24

List some key population databases used for assessing UVs (specifically SNVs/Indels).

Answer 24

• gnomAD database
  
  • 123,136 exomes and 15,496 genomes from unrelated individuals of many different ancestries.
  • Contains all ExAC data.
  • Individuals with severe paediatric disease have been excluded from the data set so can be used as a useful reference set of allele frequencies for disease studies.
  • Adult onset disease NOT excluded so must keep that in mind when using data
• 1000 genomes project - data is rolled into gnomAD
• NHLBI GO ESP Exome Variant Server (EVS)
  
  • 6503 samples from EU and African American ancestry

Question 25

What considerations must be taken into account when assessing UVs with regard to Co-occurrence with known deleterious mutation?

Answer 25

• For some dominant diseases the identification of a UV in a patient who also has a known pathogenic mutation (in trans) can provide a powerful means of excluding pathogenicity e.g. where a second pathogenic variant would be lethal.
• However it is essential to establish the phase (where possible) of the pathogenic variants and for this parental samples may be required if available.

Question 26

What considerations must be taken into account when assessing UVs with regard to Co-segregation with the disease in the family?

Answer 26

This approach is most useful as a means of excluding pathogenicity in cases where the variant does not segregate with disease.

However segregation studies can be performed in families to see if the UV is segregating with disease.

Question 27

What are the limiations of co-segregation analysis?

Answer 27

1. There is the possibility of phenocopies
2. Partial penetrance (i.e. unaffected has variant)
3. Family structure is important and non-paternity should be considered.
4. A variant that looks to segregate with disease may actually not be the pathogenic variant but be in linkage disequilibrium with it.

Question 28

What methods are available to assess co-segregation data?

Answer 28

• Møller et al (2011): segregation analysis (SISA) method: The number of informative meioses are assesed to get a probability that the variant segregated by chance. e.g. 4 meisoes = (1/2)^4 = 0.0625.
  
  • Thus, the probability that this variant is disease causing is 1-0.0625 = 0.9375 = 93.75%
  • This method can be used to combine data form different families.
• Jarvik and Browning (2016) propose guidelines on integration of co-seg into the ACMG guidelines and discuss more complex methods;
  
  • Thompson-Bayrak-Toydemir Bayes Factor method can achieve more precision when penetrance is incomplete (more complex..)

Question 29

What considerations must be taken into account when assessing UVs with regard to de novo variants?

Answer 29

• For sporadic diseases analysis of parental samples can give a strong indication as to whether a UV is pathogenic or not.
• If the UV is identified in an unaffected parent then this provides evidence that the variant is benign
• However, i_ncomplete penetrance / adult onset_ has to be considered
• It is also possible that a parent might be mosaic for the mutation and so may not be expressing the mutation in the affected tissue.

Question 30

What precautions must one take when using in silico tools for assessment of UVs?

Answer 30

• These tools have not been validated for clinical use and the outputs should be interpreted with caution.
• It has been shown that whist many tools have high sensitivity they alos have low specificity i.e. prone to false-positive results

Question 31

What is the theoretical basis of using species conservation for in silico analysis of UVs?

Answer 31

1. In a given set of homologous genes, a large fraction of amino acid sites will be conserved even among distantly-related species.
2. Variations that arose at such positions throughout evolutionary history have evidently been under strong purifying selection and eliminated from populations.
3. This suggests that the existing amino acid residues at invariant positions are critical for proper gene function.
4. Thus, information from inter-specific alignments can indicate amino acid residues that are likely to produce disease if mutated in humans.

Question 32

What are the key considerations when using species conservation/alignement for in silico analysis of UVs?

Answer 32

• When performing conservation analysis (and using protein prediction software) the quality of the alignment is key.
• Using a different alignment may result in a different result (e.g. Align GVGD is very sensitive to alignment quality, better results obtained using software’s own alignment rather than alignment in Alamut).
• The interpretation of species conservation should be cautiously applied particularly if only one species is identified for which the variant is not conserved.

Question 33

What are the best practice guidelines for using species conservation/alignement for in silico analysis of UVs?

Answer 33

• The ACGS guidelines recommend that alignments should include the full length of;
  
  • eight orthologous genes and
  • at least five mammalian orthologues plus chicken, frog and fish
• The inclusion of a species without an appropriate orthologue can result in misinterpretation of the results.

Question 34

Are multiple species alignments performed at the protein or nucleotide level?

Answer 34

• Conservation analysis is normally performed at the protein level
• However, nucleotide-sequence based predictions in non-coding and coding DNA can also be used.
• The phylogenetic scope is typically narrower than that for proteins as only a small fraction of non-coding bases align to fish genomes and there is no acceptable conservation outside vertebrates.

Question 35

What methods are avialable for performing alignments based on nucleotide sequence?

Answer 35

• GERP is a well-known method of inferring nucleotide-level constraints in genomic sequence alignments
• These methods are united by the principle of estimating observed rates of evolutionary change and contrasting these estimates with rates expected for neutral positions.
• Cooper and Shendure present evidence to suggest that n_ucleotide-level metrics are as powerful as protein-based metrics_, even without knowledge of protein biochemistry.

Question 36

How common are protein level alignment tools vs nucleotide level tools and what kinds are information are assessed in addition to simply the alignment?

Answer 36

• Nearly all protein prediction programs use alignments of homologous proteins to estimate deleteriousness (quality of alignment key to quality of results).
• Many programs also incorporate biochemical data including amino acid properties (grantham matrix), sequence information (such as presence of a binding site) and structural information (such as the presence of a beta sheet).

Question 37

What are the most commonly used protein conservation in silico tools?

Answer 37

• The most commonly used protein packages in the diagnostic laboratories are SIFT, PolyPhen and Align GVGD.
• Align GVGD combines the Grantham matrix with sequence alignments.
• SIFT and PolyPhen are both based on multiple sequence alignments and
• PolyPhen takes into account 3D data when available

Question 38

Do any tools use a combination of protein and nucleotide alignment?

Answer 38

• The Consurf server is a bioinformatics tool for estimating the evolutionary conservation of amino/nucleic acid positions in a protein/DNA/RNA
• Based on the phylogenetic relations between homologous sequences.
• Each position is given a score from 1-9 (variable to highly conserved) which should reveal those regions which are functionally relevant.
• http://consurf.tau.ac.il/2016/

Question 39

What are the consensus splice site sequences?

Answer 39

5’ end of intron = GT

3’ end of intron = AG

Question 40

What types of variants can have an impact on splicing?

Answer 40

All types of variant can impact splicing.

1. Deep Intronic,
2. Intronic (consensus splice)
3. synonymous
4. non-synonymous

Question 41

What splice prediction in silico tools are commonly used by laboratories?

Answer 41

1. Splice-site Finder-like
2. MaxEntScan
3. GeneSplicer
4. ESEfinder
5. NNSplice
6. RESCUE-ESE

Question 42

Why might RNA studies be useful for assessing UVs?

Answer 42

• RNA studies can be used to investigate the potential effect of a sequence change on normal splicing.
• This is assessed by studying cDNA generated from a fresh mRNA sample.

Question 43

What are the considerations when performing RNA studies?

Answer 43

1. Normal isoforms – may complicate the interpretation of results
2. Expression of mRNA in blood- need to consider if the mRNA is expressed in blood before offering RNA studies; another tissue may be required for RNA studies. In addition, RNA studies assume that if a mutation causes aberrant splicing in leukocyte cells it will cause aberrant splicing in tissue affected by disease.
3. Quality of RNA – degrades quickly
4. Nonsense mediated decay (NMD)-If a variant results in a splicing aberration that creates a premature termination codon the mRNA may be subject to NMD. This may mean that the variant splicing product is not present or only present at a lower level in the cDNA. Confirmation of bi-allelic expression of the variant can rule out NMD. Cannot confirm this for intronic variants.

Question 44

How might you overcome some of the typical problems encountered when performing RNA studies?

Answer 44

Some of these problems can be overcome by the use of mini-gene assays to assess splicing but these are often out of the scope of diagnostic laboratories due tot he technical skills required.

Question 45

What is a minigene assay?

Answer 45

• Minigenes provide a valuable tool for researchers evaluating splicing patterns both in vivo and in vitro
• Specifically, minigenes are used as splice reporter vectors (also called exon-trapping vectors) and act as a probe to determine which factors are important in splicing outcomes.

Question 46

How can LOH studies provide evidence for or against pathogenicity of a UV?

Answer 46

For tumour suppressor genes, LOH analysis of tumour tissue can provide evidence for or against pathogenicity.

e.g. UV has been detected in constitutional DNA and may be pathogenic,

LOH of the normal allele in tumour tissue, lindicates that the variant allele is pathogenic.

If the variant is not seen indicates that the allele with the variant has been lost – lends weight to variant not being pathogenic

Question 47

What are the limitations of performing LOH studies?

Answer 47

• Considerations: Doing LOH studies assumes that the 2nd hit is usually a large deletion – in some diseases this is more likely than others
• If LOH is seen it may be that the UV is not the causal variant but something else in cis with it is.
• Presence of normal tissue in tumour may obscure results.

Question 48

What role do functional studies have in UV assessment?

Answer 48

1. Functional studies can provide powerful support of causality for a given phenotype
2. A reliable functional assay is generally regarded as one of the best means of confirming pathogenicity.
3. Functional studies are rarely part of a routine diagnostic service due to technical difficulty/inefficiency.
4. There is a wide range of functional assays that can be performed
5. When relying on reports in the literature laboratories should be careful to scrutinise the context in which the assay is performed as the diagnostic evaluation of a functional assay can be difficult.
6. Not all functional assays are equivalent

Question 49

What kind of data can be gained from functional studies?

Answer 49

• Functional studies may consist of
• in vitro demonstration of molecular consequences (such as disruption of expression or protein misfolding)
• in vivo recapitulation of the human phenotype in a model organism.
• Functional assays require the identification of a measurable property that associates with function, which remains difficult as little is known about the function of most human proteins.

Question 50

What precautions should be taken when interpreting the rsults of functional assays?

Answer 50

• Functional studies are often not definitive.
• Negative results: need to think whether the assay was performed in the appropriate context
  
  • Genomic context (dependencies on flanking sequence or chromatin state)
  • Organismal context (functional consequences of a mutation that are species-specific)
  • Developmental context (dependency on cell type or developmental stage).
• Positive Results: does not necessarily establish causality.
  
  • e.g. a mutation may influence transcriptional regulation but it is difficult to assess if an increase in transcription is significant.

Question 51

Give an good example of a functional study aiding the interpretation of UVs in a clinical setting.

Answer 51

In a putative Lynch syndrome family carrying MSH2 and MSH6 variants of uncertain significance, functional analysis using an in vitro mismatch repair assay revealed which was the pathogenic variant (Kantelinen et al., 2011, Familial Cancer;10:515-520).

Question 52

Give an example an attempt has been made to standardise functional studies for UV assessment

Answer 52

• Crouch et al, 2008 proposed a a three-step assessment model to assess whether a UV in one of the mismatch repair (MMR) genes is pathogenic.
• Step 1 is the identification of the UV, in
• Step 2 the effect of the variant on the function of the protein is assessed in an in vitro MMR and in silico assays. According to this system, variations showing MMR deficiency in these assays indicate pathogenicity, whereas variations with no apparent MMR deficiencies require
• Step 3, a selection of biochemical assays, for further characterisation of the effect of the UV

Question 53

Give an example of how enzyme analysis can provide evidence for or against pathogenicity of a UV.

Answer 53

1. Non-ketotic Hyperglycinaemia is an autosomal recessive disorder of glycine metabolism resulting in severe neurological disease. NKH is caused by a deficiency in the glycine cleavage enzyme system. Enzyme analysis involves investigating the activity level of the glycine cleavage enzyme in liver tissue and can help determine if a UV is having an effect at the protein level.
2. Tyrosinaemia type 1 which is an autosomal recessive disorder caused by a deficiency of the enzyme fumarylacetoacetate hydrolase (FAH). Following the identification of a UV in the FAH gene enzyme analysis can be carried out to help determine the pathogenicity.

Question 54

Give a common usage for IHC in aiding interpretation of UVs.

Answer 54

• When determining the pathogenicity of a variant in the MMR genes immunohistochemical staining can provided evidence for or against pathogenicity.
• If a UV is found in a gene which shows loss of expression of the protein in tumour tissue this supports the concept that this UV is clinically significant.
• Limitations are that pathogenic missense UVs may not lead to loss of expression of a MMR protein.

Question 55

What is the basis for setting standards for UV assessment/classification?

Answer 55

• To ensure inter- and intra- laboratory consistency of interpretation UVs.
• Ensure that patient care is consistent irrespective of where they are tested.
• Becoming more important as vastly more UVs are encountered in labs adopting panel/WES/WGS technologies.

Question 56

What classification system is typically used for SNVs and who created this system?

Answer 56

• Most laboratories have adopted a classification system that consists of 5 classes

1. certainly not pathogenic
2. unlikley to be pathogenic
3. Variant of unknown significance
4. likely to be pathogenic
5. certainly pathogenic

• This was proposed by the IARC UV Working Group for the interpretation of cancer susceptibility genetic test results (Plon et al., 2008).
• Each of the classes is associated with specific recommendations for clinical management of at-risk relatives that will depend on the syndrome

Question 57

How did the Plon IARC system propose that variants were triaged into the 5 classes?

Answer 57

• The proposed classification of s equence variants was based around probability of being pathogenic, as follows;
• Class: 1 >99%
• Class 2: 0.95–0.99
• Class 3: 0.05–0.949
• Class 4: 0.001–0.049
• Class 5: <0.001
• It was proposed that a Bayesian system to generate a posterior probability should ultimately be the standard for all variants

Question 58

According to UK best practice guidelines, how should UVs be classified?

What are the current limitations of of the UK guidelines?

Answer 58

• The use of a 5 class system is considered essential for the standardisation of report wording and follow up studies.
• Whilst the guidelines set quality standards and lines of evidence that should be reviewed in assessing whether a variant might be pathogenic, the levels of pathogenicity to place a variant into any given category, and clinical recommendations associated with each category, were not included

Question 59

The Plon/IARC guidelines were devised specifically for cancer gene UVs.

What effords have been made since then to standardise UV assessment for all rare disorders?

Answer 59

In 2015 the American College of Medical Genetics and Genomics and the Association for Molecular Pathology published a comprehensive set of “Standards and guidelines for the interpretation of sequence variants”

Commonly known as the “ACMG” guidelines

Question 60

What are the defining features of the ACMG guidelines?

Answer 60

• They also endorse a 5-class system but with different probability thresholds of >90% (class4) and <10% (class2).
• Importantly the guidelines define specific pieces of evidence that can be used in support of pathogenicity or lack thereof.
• They then outline how the combination of pieces of evidence can place a variant in a particular variant classification category.
• Pieces of evidence may be considered strong, moderate, supporting, weak etc

Question 61

How is the ACGS changing its guidelines to relect this new approach to UV classification?

Answer 61

• In Nov, 2016 the ACGS issued a “consensus statement on adoption of ACMG guidelines for sequence variant classification and interpretation”
• ACGS currently in close collaboration with the ACMG to issue updated UK guidelines which used ACMG as the base and provides addtional information to clarify ambiguities in the original ACMG document
• An ACGS variant classification working group is currently circulating variants accross laboratories in the UK to facilitate the production of the new guidelines.