Variant Classification Webinar Series Q&A

Webinar One (Gathering the Pieces: Tools for Variant Classification – 11/12/15) Questions and Answers:
Question
: Is VAAST a useful in silico prediction tool?
Answer:  VAAST is one of a number of next-generation in-silico scores. Published evidence suggests that it is superior to routinely used scores such as SIFT and Polyphen. We do not rely heavily on any of the in silico scores, however, as all of them are prone to both false negatives and false positives. Per the ACMGG guidelines, they are treated jointly as supporting evidence.

Question: Are we going to discuss the relative value of the in silico algorithms?
Answer: In silico tools are considered as a whole entity for supporting evidence and not considered individually.

Question: How do you find institutions or labs to perform functional studies for that particular VUS?
Answer: Depending on the gene, we may have active research collaborations with scientists outside of GeneDx.  However, we do not routinely have access to functional in vitro/in vivo studies for our variants.

Question: Can truncated protein induce gain of function?
Answer: The most common effect of protein truncation is to fail to produce a protein due to non-mediated-decay.  If a sufficiently distal truncating event occurred that resulted in a translated protein product, a gain of function could occur but is infrequent.

Question: In cases of in-frame deletions, how can you predict the result on the protein and thus, the pathogenicity of a variant?
Answer: We currently predict it based on the relevance of the region deleted (conservation, functional domain residues), publications, and/or similar mutations.

Question: How general are the ACMG guidelines for non-Mendelian disorders? Are the principles the same for interpreting variants in complex genetic traits? What are the cautions for those traits?
Answer: The current guidelines are not designed for classification of non-Mendelian disorders (mitochondrial, somatic) or complex traits. Different sets of interpretation guidelines need to be developed to consistently classify variants for these disorders.

Question: What do you think of the Leiden database for that or in general?
Answer: Most of the LOVD gene databases are very good and contain valuable information. The use of individual databases for a particular gene is approved by our in-house gene experts.

Question: I have a little girl with a FLNC mutation – she has myopathy with partial vent dependence – biopsy non-specific – what dbs ideal?
Answer: The only database that we are familiar with for FLNC is the one hosted by LOVD. Additional resources that could be used to determine pathogenicity of this variant are HGMD to determine if this variant is published and view nearby published variants, UCSC to determine conservation of the variant across species, and various in silico prediction models.

Question: Are you able to deposit in ClinVar the phenotype in which a variant was found?
Answer:  Our current data is expressed in ClinVar at the variant level and represents an aggregate of all case information that is readily available. We are working toward a case-level system long term that will allow for individual phenotypes of the patients to be viewed if that information is available under the “supporting observations” tab; however, our current internal system is still in development for this process. Once this system is in place, we will aim to provide as much clinical information as possible. It should be noted that we still do not receive phenotypic information for a large portion of our cases, so this may not be available for all variants if the data have not been provided to us.

Question: When I look at ClinVar at one of your classifications, when I click the link for supporting data most often I get a copy of your lab’s SOP rather than verifiable data. Is this a stopgap or the plan moving forward? Will verifiable data ever be included in ClinVar (eg. Phenotypes, pedigrees, etc)
Answer: The summary evidence tab information is our standard reporting language summarizing our variant classification assessment.  Our current data are expressed in ClinVar at the variant level and represents an aggregate of all case information that is readily available. We are working toward a case-level system long term that will allow for individual phenotypes of the patients to be observed if that information is available under the “supporting observations” tab ; however, our current internal system is still in development for this process. Once this system is in place, we will aim to provide as much clinical information as possible. It should be noted that we still do not receive phenotypic information for a large portion of our cases, so this may not be available for all variants. With regards to pedigrees, most of these are also not provided or stored in an easily extractable manner. ClinVar currently does not accept pedigree relationship information in the formal sense; however, there are initiatives to provide this information long term. Please contact Lisa Vincent (lvincent@genedx.com) for additional information regarding these potential initiatives and how your team could also participate.

Webinar Two (Putting the Pieces Together: Applying the ACMG Guidelines to Variant Interpretation – 12/3/15) Questions and Answers:
Question: How do you determine the cutoff for the ‘absent in large population databases’?
Answer: At GeneDx, we take into consideration the data quality and source of each population data set.  As indicated in the ACMG guidelines, if a variant appears to be completely absent, then it is important to confirm that there is sufficient read depth at that position to know that data is reliable.  A variant can appear to be absent if there is not sufficient coverage at that position.  Additionally, there can be random errors and/or artifacts in the sequencing data in some of the large publically available data sources.  We have internally set cutoffs based on the issues specific to each dataset for what we consider “absent”.  We take into consideration the aggregate of data available to us.  For example, if a variant is absent in our internal dataset and 1000 Genomes but was reported in 1 individual in ESP, we would still apply PM2 criteria.

Question: We struggle with interpreting synonymous variants with no predicted impact on splicing, but are present in very low frequency in population databases and/or have not been seen by other laboratories. While we suspect that these variants are not clinically significant, the new guidelines make that a difficult classification to make.
Answer: We agree that this scenario is challenging and an adjustment for all the labs using the new guidelines.

Question: Do you ever pool or meta-analyze data from published studies? For example, you stated that the 2 case-control studies in the cancer example had odds ratios that were not significant. But I wonder if meta-analyzing them would result in an odds ratio that was statistically significant.
Answer: Pooling studies can often be difficult since they may be using different populations and/or have different confounding variables.