Whole genome sequencing; not ready for prime time
A test of the utility of whole-genome sequencing (WGS) on 12 genetically healthy adults reveals several shortcomings, according to a study published in the March 12 issue of JAMA.
The report describes "the process that a team at an academic medical center went through to conduct sequencing from beginning to end. They raise a lot of interesting practical issues that have been circling for a while and provide preliminary numbers to quantify the process," Cinnamon S. Bloss, PhD, director of social sciences and bioethics at Scripps Translational Science Institute, La Jolla, California, told Medscape Medical News.
As costs plunge and time shrinks for sequencing human genomes, clinicians are considering how to use DNA data in their practices. Thus far, exome sequencing and limited WGS have performed well for individuals with atypical presentations or undiagnosed diseases, as well as for specific groups, such as children with developmental delay or admission to a neonatal intensive care unit. However, little is known about the technique's value in healthy adults.
Probing a Dozen Average Genomes
Therefore, Frederick E. Dewey, MD, from the Stanford Center for Inherited Cardiovascular Disease, Stanford Cardiovascular Institute, Stanford Center for Genomics and Personalized Medicine, and Division of Cardiovascular Medicine, Stanford University, California, and colleagues evaluated genomes from healthy adult volunteers at their center. The investigation considered coverage of clinically relevant gene variants, concordance using different sequencing platforms, measurement of risks for diseases and drug responses, resources required to evaluate data, and clinical follow-up.
They used an Illumina platform for sequencing, with confirmatory sequencing for 9 participants on the Complete Genomics platform. The researchers used appropriate reference genomes for the 5 white European and 7 East Asian participants and screened genes from many sources, including the Human Gene Mutation Database (HGMD), 56 genes that the American College of Medical Genetics and Genomics recommended for prioritization in testing because of actionability, 118 gene variants associated with diabetes mellitus type 2 or coronary artery disease, and 555 genetic variants associated with drug responses.
A team of 3 genetic counselors, 3 physician informaticists, and 1 molecular pathologist curated sequence variants from literature reviews and phenotype identification software that uses allele frequencies, functional classes, and evolutionary conservation to predict pathogenicity. Team members determined which variants to report to 5 physicians, who then suggested clinical follow-up solely on the basis of the DNA data. Three were academic primary care practitioners, 2 of whom had no experience with genetics or genomics, and the other 2 were academic medical geneticists.
Omissions and Uncertainties
Sequencing a genome requires randomly cutting many copies of the genome and then piecing back together overlapping sequences. The more copies of the genome that are sequenced (referred to as coverage depth), the more of the genome appears in the completed WGS. For these dozen patients, sequencing missed 10% to 19% of disease genes, depending on database and platform.
However, with the addition of targeted sequencing, detection of known variants in disease-causing genes and drug response variants was 99% to 100%, indicating that it is much easier to find known targets. The platforms were also better at detecting single-nucleotide variants (99% - 100%) than small insertions and deletions (53% - 59%).
The software identified 90 to 127 "novel and rare" genetic variants and up to 4 large structural variants per participant. Database and literature review for evidence of pathogenicity took a median of 54 minutes (5 - 223 minutes) per variant, at an estimated median cost for sequencing plus interpretation of $14,815 ($14,050 - $15,715).
Interpretation Varies, Even Among Experts
The investigators randomly selected 18 genetic variants to assess how likely 6 genomics professionals were to agree on potential personal risk or carrier status. The results indicated moderate interrater agreement on pathogenicity (Gross κ, 0.52; 95% confidence interval [CI], 0.40 - 0.64) and fair interrater agreement on suitability for reporting (Gross κ, 0.83; 95% CI, 0.73 - 0.93).
The researchers downgraded disease associations for some mutations based on context, such as whether a variant is pathogenic in a particular population. The team found 47 of 68 variants (69%) listed in HGMD as "disease-causing" to be less severe or variants of uncertain significance.
For each participant, WGS revealed 2 to 6 "personal disease-risk findings," most for adult-onset conditions, from the HGMD. Each participant carried 8 to 18 disease-causing recessive alleles. For the American College of Medical Genetics and Genomics–reportable genes, each participant had 12 to 20 variants, which fell to 1 to 7 variants after eliminating sequencing artifacts or common polymorphisms.
Overall, the study identified only a single "very likely pathogenic" gene variant, a 19-base deletion in BRCA1. The woman, who did not have a family history of the associated cancers, had risk-reducing surgery. Eleven of the 12 participants had at least 1 gene variant of pharmacogenetic significance.
The physicians suggested 1 to 3 diagnostic tests and/or referrals per participant, with a median cost of $351 to $776, which include a "high-complexity established patient visit" and 2 hours of genetic counseling.
The researchers conclude that in their study, WGS "was associated with incomplete coverage of inherited disease genes, low reproducibility of detection of genetic variation with the highest potential clinical effects, and uncertainty about clinically reportable findings."
Limitations of the study include the early stage of applying sequencing data to clinical decision-making, lack of full clinical information on the participants, and applicability beyond academic medicine.
In an accompanying editorial, William Gregory Feero, MD, PhD, from the Maine Dartmouth Family Medicine Residency, Augusta, and associate editor, JAMA, acknowledges that although medical decision-making with incomplete understanding of underlying biology is ancient, clinicians should "recognize that determining the biological consequences of very rare and novel variations" in a genome "remains arduous, tenuous, and costly even in highly experienced hands."
Dr. Bloss noted some trade-offs in the economics of WGS. "You might order a more expensive test if it's the most cost-effective and shortest road to diagnosis," she said, mentioning the diagnostic odysseys common in the rare disease world. "In those cases, clinical sequencing has the potential to be very cost-effective and avoid downstream tests. But I think we are a long way from seeing this type of analysis be cost-effective for someone who is relatively healthy. We'd all like to see the utility of WGS for disease prevention, but I don't know that we're there yet."
Dr. Dewey is a stockholder and member of the scientific advisory board of Personalis Inc and receives royalties for patents related to genome sequencing. One coauthor has received speaker's fees from Illumina Inc. One coauthor is on the board of Coriell Inc. Three coauthors receive royalties for patents related to genome sequencing. Four coauthors are founders, stockholders, and members of the board of Personalis Inc and receive royalties for patents related to genome sequencing. One coauthor is a stockholder and member of the board of NuMedii Inc; consultant to Lilly, Regeneron, Johnson & Johnson, Roche, Geisinger, Verinata, Pfizer, and Samsung; has received speaker's fees from Pfizer, Lilly, Siemens, Bristol-Myers Squibb, and Genentech; and holds stock in Carmenta, Eceos, Assay Depot, and Genstruct/Selva. One coauthor is on the board of and owns stock in Genapsys Inc. One coauthor is a member of the board of Aviir Inc. Dr. Feero is a contributing editor for the JAMA. Dr. Bloss has disclosed no relevant financial relationships.
JAMA. 2014;311:1117-1119, 1035-1044.