Jeffrey Perkel has been a scientific writer and editor since 2000. He holds a PhD in Cell and Molecular Biology from the University of Pennsylvania, and did postdoctoral work at the University of Pennsylvania and at Harvard Medical School.
Several weeks ago, an infant in the neonatal intensive care unit at Children’s Mercy Hospital in Kansas City, Mo., came to the attention of Stephen Kingsmore.
Kingsmore is executive director of the hospital’s Center for Pediatric Medicine, and over the past several years, his lab has developed and refined methods for rapid genetic diagnosis of pediatric disorders using next-generation DNA sequencing technology. Using a high-speed run mode on the Illumina HiSeq 2500 and some clever computational algorithms that enabled geneticists to focus a variant search based on the patient’s symptoms, his team showed in 2012 that it could return a genetic diagnosis from whole-genome sequencing in as little as 50 hours after sample receipt.
Now he was being asked to apply that method to a child who had to be kept on a ventilator, because he stopped breathing whenever he fell asleep. The condition is called “central hypoventilation syndrome,” or “Ondine’s curse,” and according to Kingsmore, it is associated with multiple genes. The question was, which one?
To find out, the case was referred to Kingsmore. The child was enrolled on Friday afternoon, but samples weren’t processed until the following Monday. By Wednesday, the sequencer had completed its run; at that point, automated software tools took over, identifying and ranking genetic variants based on their likely relationship to the disease. Thursday morning, when Kingsmore came into work, the data were ready for analysis. “Within a few minutes, I had the diagnosis,” he says. The software had flagged a mutation in one of the genes initially deemed unlikely to be associated with the disease—“It was ranked No. 99 in terms of likelihood of causing the baby’s features,” Kingsmore notes—and the mutation’s identification could help guide the child’s treatment, or at least clarify difficult choices.
Typically, until clinicians arrive at a diagnosis, the “end-of-life decisions or surgical decisions, the big decisions, are on hold,” he says.
Using next-generation sequencing (NGS) to arrive at those diagnoses is increasingly common in today’s clinics. Powered by sequencers from Illumina, Thermo Fisher Scientific and Pacific Biosciences, clinical labs are churning out DNA sequence data in ever greater numbers. As a result, researchers and clinical geneticists are finding it easier than ever to translate medical mysteries into clinical action.
Of panels, exomes and genomes
Generally speaking, genetics labs run three types of tests: targeted gene panels, exomes and whole human genomes. Panels, comprising anywhere from one to hundreds of genes, are typically the least expensive to perform and simplest to interpret, but they are only informative if a key mutation is in the tested genes. Human exomes focus on the genome’s protein-coding sequences, making them highly interpretable but also somewhat limited, as many mutations fall outside the covered regions. Whole human genomes read every base, providing the most complete data but also the most challenging analyses—not to mention the highest prices. As a general rule, the less you sequence, the deeper you can read, notes Niall Lennon, director of clinical development at the Broad Institute of Massachusetts Institute of Technology (MIT) and Harvard, Clinical Research Sequencing Platform. “Oftentimes, you will find cancer panels sequenced to a 500x or 1,000x mean. That’s actually much easier in a panel than an exome.”
Thermo Fisher Scientific offers two oncology panels in the United States, for research use only, says Mike Nolan, the company’s vice president and oncology group general manager. Capturing either 52 or 143 genes, the tests require just 10 ng of input DNA, Nolan says—a limit that is compatible with precious biopsies. According to Nolan, the specific genes on those panels were selected in partnership with pharmaceutical companies to help inform treatment decisions, testing for mutations that may influence drug selection and clinical trial participation. The TaGSCAN panel at Children’s Mercy Hospital in Kansas City tests some 514 genes associated with childhood disease, including genes for “neurodevelopmental, metabolic, nuclear mitochondrial, intellectual disability and immunodeficiency conditions.”
In the late 2000s, says David Dimmock, one of the clinical directors of the genetic testing laboratory at the Medical College of Wisconsin (MCW), “there was an evolution towards gene panels, [where] you could say, these six or 10 genes account for a good portion of this disease.” More recently, however, researchers have increasingly recognized that the set of genes that can cause a given disease is much wider than previously recognized. As a result, exomes have become increasingly popular, he says. “If we [sequence] everything in one go, we may get a surprise we can link back to phenotype, and that’s better than trying to test every gene first.”
The Broad Institute has some “40 to 50” Illumina instruments, Lennon says, including HiSeq X Tens, HiSeq 4000s and a dozen HiSeq 2500s, the latter of which have been validated for clinical work. But the vast majority of the lab’s work is for basic research, he says: “We produce a 30x whole human genome every 30 minutes.” In June 2015, the facility pumped out nearly 260 terabases of data—“8.6 trillion bases per day.” Yet the lab’s clinical output is just 30 to 40 exomes per week, compared with “thousands” per week on the research side.
According to Lennon, the lab offers two flavors of exome sequencing: germline exome sequencing (at 100x depth) and cancer somatic exome sequencing (150x depth). Turnaround time is 21 days from sample receipt to report, he says. “The types of projects that use exomes here are in the middle,” Lennon says. “They want to make a diagnosis but also to discover or look at nonobvious genes. I also see cases where people go for a panel first and then [pivot] to an exome, if they see nothing.”
But critically, the Broad does not interpret its exomes, Lennon says. “We deliver raw data, variants and in some cases annotations. But that’s where we stop.” So it’s up to the researcher to interpret the data. In contrast, both MCW and Children’s Mercy Hospital provide interpretation as part of their services, and both do so in more or less the same way: After the sample is sequenced, the sequences are aligned to a reference genome for variant calling. Software tools then rank those variants based on how likely they are to affect a protein, whether they have been associated with disease in the past, allele frequencies and so on, to produce a list of likely candidates. At that point, follow-up studies may be required. MCW, for instance, confirms likely variants by Sanger sequencing prior to reporting, says Dimmock. “That actually is the longest part of the process.” (The lab can turn around a rapid exome in two weeks.)
A 2014 study in Science Translational Medicine by the Kingsmore lab demonstrates NGS’s potential power in the clinic [1]. The lab tested 100 families with 119 children affected by neurodevelopmental disorders, using either whole-genome or whole-exome sequencing of parent-child trios. “Forty-five percent received molecular diagnoses,” the authors found: 11 (73%) of 15 families tested by whole-genome sequencing, and 34 (40%) of 85 families tested by whole-exome sequencing (in one case, “staged” exome sequencing followed by whole-genome sequencing). On average, the cost of the diagnostic journey prior to sequencing was $19,100 per family, which means sequencing would be cost-effective, the authors wrote, “as long as the cost was no more than $2,996 per individual.”
Clinical toolkit
Generally speaking, any sequencer can be applied to clinical genetics. But some are designed with that application in mind, including Illumina’s MiSeqDx and Thermo’s Ion Torrent PGM Dx. Pacific Biosciences does not yet have a clinical sequencer—its systems are intended for research use only—but it is collaborating with Roche to bring its long-read chemistry into the diagnostics space, says Jonas Korlach, chief scientific officer at Pacific Biosciences. Long-read chemistries could offer certain advantages for clinical applications, Korlach says. These include the ability to monitor transcript isoforms and mutation phasing (that is, the presence of multiple mutations in a single transcript as opposed to single mutations on multiple transcripts) as well as large-scale structural variation. In one study, for instance, researchers leveraged PacBio’s multi-kilobase reads to sequence full-length BCR-ABL1 transcripts, data that could be used in the future for earlier prediction of tumor relapse.
Also available for clinical genomics are target selection systems for enrichment of gene panels or exomes, such as Agilent Technologies’ SureSelect and Roche NimbleGen’s SeqCap EZ reagents. Agilent, for example, recently launched its OneSeq Target Enrichment kit for identification of copy-number variants, as well as a series of cancer-associated gene panels called ClearSeq.
Such tools make clinical sequencing easier than ever. Yet the technique remains something of an “esoteric diagnostic,” Lennon says. In part, that’s because of a lack of hard data on efficacy. Few studies like Kingsmore’s have been run, and all have been either retrospective analyses or prospective case studies (essentially, n-of-1 studies). What’s missing, Kingsmore says, are prospective, randomized studies on cost-effectiveness and clinical utility—proof that improved diagnostic rates translate into clinical benefit. “Payors as yet are unconvinced.”
Meanwhile, the technology continues to evolve. Kingsmore says his team has already cut the time of its rapid whole-genome sequencing test from 50 hours to 26, thanks to new computer algorithms and sequencer improvements. “It’s a much better test now than it was in 2012, it’s much more sensitive.” For patients, parents and physicians everywhere, that should make difficult medical decisions that much easier—and faster—to make.
References
[1] Soden, SE, et al., “Effectiveness of exome and genome sequencing guided by acuity of illness for diagnosis of neurodevelopmental disorders,” Sci Transl Med, 6:265ra168, 2014. [PubMed ID: 25473036]
[2] Cavelier, L, et al., “Clonal distribution of BCR-ABL1 mutations and splice isoforms by single-molecule long-read RNA sequencing,” BMC Cancer, 15:45, 2015. [PubMed ID: 25880391]
Image: Shutterstock