NEW YORK (GenomeWeb) – A new study by researchers from the Baylor-Hopkins Center for Mendelian Genomics has provided data that the authors hope will inform ongoing discussion on how best to interpret and return to patients incidental findings from genomic sequencing.
In the study, published last week in Genetics in Medicine, only two of more than 200 patients from the Baylor-Hopkins research program had a returnable incidental finding.
The American College of Medical Genetics recommended in 2013 that clinicians report to patients all pathogenic and likely pathogenic variants in 56 genes when patients receive whole-exome or whole-genome sequencing.
While the ACMG's guidelines apply to clinical sequencing, ongoing research studies that are returning results of whole-exome and whole-genome sequencing to patients — such as the National Human Genome Research Institute's Clinical Sequencing Exploratory Research program — have also begun to grapple with the ethics and practicalities of interpreting and returning incidental findings.
In mid-2014 members of the Clinical Sequencing Exploratory Research consortium's return of results committees, along with the Electronic Medical Records and Genomics Network Consent, Education, Regulation, and Consultation working group published their own recommendations intended to help researchers navigate the issue, concluding that research sequencing programs need not specifically seek out pathogenic variants in the ACMG's 56 genes, which can require considerable extra time and effort, but should report them to consenting subjects if discovered incidentally.
The Baylor-Hopkins study began soon after the ACMG's recommendation while the field was in the midst of weighing the guidelines in the context of research. The group decided to use a cohort of exome-sequenced patients to look at the prevalence of alterations in the ACMG's list of 56 genes, and to evaluate different strategies for determining their pathogenicity and whether or not they should be reported to subjects.
According to the authors, the study illustrates both the potential and current challenges inherent in assessing the phenotypic or clinical consequences of incidental variants in whole-exome data.
In the study, the team analyzed whole exomes from 232 individuals sequenced as part of the Baylor-Hopkins NHGRI-funded program, which is focused on diagnosing Mendelian disorders. The researchers reanalyzed the sequencing data for these patients, specifically looking for rare splicing or non-synonymous SNVs and indels within the 56 ACMG-recommended genes.
Overall, the team identified 249 variants, some of which occurred in as many as seven different subjects, resulting in 391 total variants across the 232 individuals, or an average of 1.69 variants per patient. Only 45 subjects showed no variants in the 56 genes.
The researchers then stratified these discovered variants by type. Most — 231 of 249 — were missensse mutations. There were also four non-synonymous splice variants, two other splice variants, three frameshift indels, four non-frameshift indels, and four nonsense variants.
Julie Jurgens, the study's first author, told GenomeWeb that one main goal for the group was to investigate how these different variants might stratify using different criteria for evaluating their pathogenicity or reportability.
First, the group looked at the classification of the variants by three available databases, the Human Gene Mutation Database (HGMD), ClinVar, and a database fromEmoryUniversitythat has since been integrated into ClinVar. Among the 249 alterations, about half were classified by at least one of the three databases. However, almost half of the variants were listed discordantly by the different datasets, the authors wrote.
The team then went a step further, applying a technique for filtering reportable variants developed and described by another group, the National Heart, Lung, and Blood Institute Exome Sequencing Project.
"We thought a lot about whether we should decide our own criteria to define what was pathogenic or likely pathogenic, or benign or likely benign, or of unknown significance," Nara Sobreira, the study's senior author, told GenomeWeb. "Then this paper came out and we thought they did a good job. We didn't think we would be able to come up with something better, and we felt that by applying their criteria we could have a better idea of where it could potentially be improved."
The NHLBI method incorporates a variety of factors in assessing variant pathogenicity, including comparing minor allele frequencies to the reported incidences of associated phenotypes. Applying these criteria to the 391 total variants in the study, the Baylor-Hopkins researchers narrowed the field down to only two pathogenic variants and three likely-pathogenic variants. Of these, one was eliminated because it did not fit the ACMG guidelines, and another two because their associated phenotype overlapped with that of their Mendelian disorder. That left two patients who were deemed to have a reportable incidental result — a little less than 1 percent of the total cohort.
"Initially we were wondering if we could form a strategy based on the existing … HGMD, ClinVar and Emory database[s]," Jurgens said. The Emory data and ClinVar both showed a high concordance with what the group later defined as pathogenic using the NHLBI approach, she added, but the databases' overall discordance amongst themselves — as well as the fact that a full half of the variants discovered by the group did not appear in any of the databases — led the researchers to conclude that other tools would be required.
"A database can be used as one of the tools to define pathogenicity but it's a problem that very few of the variants are in these databases. Half of ours were not in any database," Sobreira said. "We decided that trying to apply these criteria that were already published would give us a good idea of how this works, and how you might need to change things to do a better job."
Another goal, Jurgens said, was to see if the group's results in terms of the frequency of reportable variants matched up with what other groups have found in separate sequencing cohorts.
Indeed, the approximately 1 percent prevalence of reportable variants tracked fairly closely to other findings, including in the NHLBI sequencing group's own work, which found about 1.4 percent of 1,000 individuals had a pathogenic or likely-pathogenic variant.
Another study published by researchers at the NHGRI analyzed variants in 37 genes in the exomes of 572 participants in the ClinSeq study and found clinically important mutations in eight, or about 1.4 percent, of them.
However, Sobreira said, these numbers are clearly dependent on the methodology used to define pathogenicity. For example, another study of stored sequencing results from the NIH's Undiagnosed Diseases Program found that a full 5 percent of patients had an incidental variant in one of the ACMG genes.
According to Sobreira, the team's results have already informed discussions within the Baylor-Hopkins group in terms of how to analyze and report incidental variants as part of the ongoing Mendelian Genomics Center research program.
She said the team has been weighing whether to more actively search for incidental pathogenic variants in the ACMG's gene list, or to lean more toward the recommendations from the CSER and eMERGE groups that research programs need not seek out variants unless doing so is part of their research protocol.
The group is leaning toward actively searching for pathogenic variants, but limiting this effort to only those that are recorded as pathogenic or likely to be pathogenic in the ClinVar database, which would greatly reduce the extra effort and time involved.