Guidelines. High throughput sequencing generates an unprecedented amount of genetic data. Most exomes are generated in a specific context, i.e. the genetic data is screened for variations in specific candidate genes or screened for de novo mutations. However, these approaches only use a small fraction of the genetic data generated per individual. High-throughput sequencing may also reveal clues towards possibly relevant diseases, and there is an ongoing debate if and how incidental findings should be returned to individuals undergoing high-throughput sequencing. Now, a recent paper in the American Journal of Human Genetics uses a very stringent clinical approach to assess the frequency of medically actionable findings in exome data. The results are not what you would think, and there is an urgent need to fix the existing databases.
Exome emergency. Is there such as thing as an exome emergency? Could you imagine any scenario where you need to call a patient at 3:00AM to schedule an immediate visit after looking through exome data? This provocative question already suggests that the urgency of returning genetic incidental findings is different from other medically relevant incidental findings, for example a brain tumor found on a research MRI. Nevertheless, learning about a predisposition for an underlying disease that might manifest in the future may be in the proband’s interest, especially if options for treatment or disease modification through early detection are available.
Medically actionable. In the study by Dorschner and colleagues, a medically actionable genetic variant has sufficient penetrance for a disease that would result in specific medical recommendations to improve outcome in terms of mortality or the avoidance of significant morbidity. For example, in patients with a genetic form of hypercholesterolemia due to mutations in LDLR, treatments are available to mitigate the effects of high cholesterol, which may include heart attack and stroke. For several genetic cardiac arrhythmia syndromes, early detection and pacemaker implantation may be lifesaving. Accordingly, genes involved in these conditions may be worthwhile looking at. The American College of Medical Genetics and Genomics has issued a specific list of genes that should be included in an assessment of genomic data. However, there are a large number of variants in these genes with limited evidence. Assessing this evidence was one of the main tasks of Dorschner and colleagues.
Pathogenic variants. The authors screened 1000 randomly selected probands from the Exome Variant Server for variants in 114 candidate genes that were considered medically actionable. Subsequently, they compared the identified variants to the Human Gene Mutation Database (HGMD), a repository that collects and catalogues all reported pathogenic variants in genes causing human disease. In total, they identified 239 unique variants. This list of variants was then evaluated by a panel of experts, with a combined experience of 340 clinical years, who spent 23 minutes on average on each variant. Variants were classified as pathogenic, likely pathogenic, variants of unknown significance or likely benign variants of unknown significance. The classification of variants was very conservative with a clinical focus in mind. Only variants that had a very low allele frequency and that had been found in several affected individuals previously were considered pathogenic. A variant reported in fewer than three affected individuals or without segregation data was not considered likely pathogenic or pathogenic. The results of applying this filter to the 239 variants was surprising.
Database noise. Roughly 30% of variants reported in HGMD were more frequent in the exome data of 6500 individuals than the disease frequency. This is a very strong argument that a specific variant is not pathogenic, but rather a chance finding. With many of the remaining variants classified as variants of unknown significance, only 16 autosomal dominant variants in 17 individuals and one individual with two recessive mutations were identified. In addition, five additional variants were identified that disrupted a disease-gene through a truncating mutation early in the protein sequence that would result in nonsense-mediated decay. In total 23 participants had variants that were considered actionable.
The diseases. The 23 probands with medically actionable findings had essentially three types of genetic diseases. The group of conditions included cancer syndromes such as hereditary breast and ovarian cancer due to mutations in BRCA1 and BRCA2; genetic cardiac syndromes due to mutations in SCN5A, KCNQ1, CACNB2 or TNNT2; or metabolic syndromes such as familial hypercholesterolemia with mutations in LDLR. Roughly 2% of individuals carried such actionable mutations, and the frequency was higher in individuals of Caucasian descent compared to African American individuals.
Not for children. The list of 114 genes selected by Dorschner and colleagues is for diseases manifesting in adulthood. For a pediatric setting, such a project might be infinitely more difficult due to the range of disease onsets in many pediatric genetic syndromes, including the epilepsies and other neurogenetic disorders. The study by Dorschner and colleagues demonstrates the need for manual review of causative variants because available, quality-controlled databases still contain a large number of variants of unknown significance. Only roughly 10% of variants in such databases were eventually classified as pathogenic or likely pathogenic. Nevertheless, a frequency of 2% of individuals with medically actionable findings is relatively high. The upcoming years will show us whether such findings can be translated into clinical care.
Lessons for EuroEPINOMICS. When exomes were first investigated, the sheer amount of information made it possible for us to find the needle in the haystack. Ever since then, we have learned a lot about rare population variants and the difficulty of reference databases to include these variant. For example, when Nobel Prize Laureate James Watson’s genome was analyzed, he was found to be homozygous for a reported mutation for Cockayne syndrome. This variant was probably misreported initially and now populates available disease databases. It will become the task of the epilepsy genetics community to provide quality-controlled databases for epilepsy genetics that will help clinicians distinguish signal from noise.