|Next-Gen Sequencing: Making a Difference in Patient Care|
An SLAS Electronic Laboratory Neighborhood September 2011 feature on next-generation sequencing (NGS) focused on the promise and challenges of using the technology to hasten the pace of drug discovery. NGS also has moved forward on the clinical front to aid in diagnosis, prognosis and therapeutic decision making.
Eric J. Duncavage, M.D., Assistant Professor of Pathology & Immunology, Anatomic & Molecular Pathology at Washington University School of Medicine in St. Louis (WUSTL) and colleagues at GPS@WUSTL are among a handful of research teams who have developed NGS DNA sequencing panels for cancer to help oncologists improve patient care. Although the WUSTL panel and others are already in use, a number of obstacles need to be overcome before NGS can reach its full potential in the clinic.
Drug Discovery versus Clinical Applications
“The members of our department realized several years ago that it was going to be impossible to keep up with the rapid pace of molecular oncology discovery using the old paradigm of single-gene testing,” says Duncavage. “We knew the only way to translate these discoveries into improved patient outcomes was to use the same NGS technologies by which they were initially identified. Therefore, the premise of our clinical NGS service is that we apply discoveries made in large-scale genomic studies to our CAP/CLIA (College of American Pathologists/Clinical Laboratory Improvement Act)-validated clinical laboratory and make them available to oncologists in a rapid and cost-effective manner. The challenge has been to apply clinical laboratory rigor to a technology that traditionally has been used for discovery, recognizing that there are a number of differences between these two areas of research.”
Those differences include goals—clinical utility versus drug discovery; quality metrics—e.g., sensitivity, specificity, positive predictive value for the clinical laboratory versus the parameters of the specific scientific question discovery is trying to solve; regulatory requirements—CAP/CLIA, HIPPA, IRB (internal review board) for the clinic versus IRB approval alone for discovery; and funding—reimbursement from insurance companies and other payers for clinical applications versus research grants for basic science.
Finding Mutations that Matter
The different aims and requirements of NGS for discovery versus the clinic have implications for the ways in which the technology is implemented, Duncavage explains. “Unlike research scientists, we don’t sequence full genomes; we leverage existing research knowledge and apply it to clinical practice. We’re trying to identify mutations that have known prognostic significance in patients—and we want the genes to be targetable by existing drugs to help clinicians in therapeutic decision making.
“Historically, genetic testing has been used mainly for constitutional disorders,” Duncavage continues. “Mutations for cystic fibrosis or Von Willebrand disease, for example, are fairly easy to detect because they involve single nucleotide variants—an A to a C or a T to a G. There are two copies of each chromosome, and one or both will be mutated, which means allele frequencies are either going to be zero, 50 or 100 percent.
“Cancer, by contrast, is much more heterogeneous and has a high degree of structural variation. In addition to single nucleotide changes, we need to detect other types of mutations, such as insertions and deletions, translocations and copy-number variation.”
Further complicating the picture is that a tissue specimen from a tumor consists of both normal and cancerous cells, with the latter making up only a small percentage of the specimen. “Specimens also contain the surrounding stroma, fibrotic cells, infiltrating lymphocytes and other inflammatory cells,” Duncavage says. “Therefore, the observed allele frequencies for these mutations are much lower than for constitutional disorders, and one of our challenges is being able to detect the lower frequency variants.”
One way to do that is by increasing sequence coverage, he adds. High coverage—that is, sequencing the same nucleotide base multiple times—increases the likelihood that as many cancer cells as possible are detected. Coverage depth is expressed as the number of times a base is read, followed by an “X.” “For discovery, the genome typically is sequenced an average of 20 to 30 times. But when we sequence in the clinical lab, we aim for 500- to 1,000-fold coverage, so for each position we sample at a much greater frequency,” Duncavage explains. High coverage also has the clinical advantages of reducing false positive/false negative rates while raising the positive predictive value of the genetic test.
High coverage sequencing is possible now because the cost of NGS is so low. “If we want to sequence for mutations in multiple genes, it’s much more cost-effective to do next generation sequencing. We can cover a huge area—500 kilobases or so—for the same price as three or four kilobases by Sanger sequencing, for example,” says Duncavage. “NGS is also more sensitive. In general, with Sanger, we'd find only mutations that are present at about a 20 percent observed allele frequency, which means that mutated cells would have to make up at least 40 percent of the sample that is being sequenced. But NGS methods will pick up mutations that are present at a five-to-10 percent observed allele frequency.”
These advantages led Duncavage and colleagues to design, validate and implement a multi-gene NGS panel at Washington University that targets 27 genes frequently mutated in leukemia and a variety of solid tumors. The target region for each gene on the oncology panel is about 500 kilobases, with 1,000X depth of coverage.
“What we test for is driven by clinicians,” explains Duncavage. “Generally, we get samples from patients with high-stage, refractory disease—those who have failed conventional first- or second-line chemotherapy or who have extensive metastatic disease, with very long odds of survival in the short term. Lung cancer is one of the most common cancers we test. If we find mutations that are targetable, then the clinician can develop highly patient-focused treatments for these individuals.”
The CAP/CLIA Difference
When investigators sequence a patient’s genome for research purposes—a clinical trial, for example—they can’t act on the information they obtain by entering it into the patient's medical record or to assist in treatment decisions because study participants must be anonymized, Duncavage observes. “With our gene panel, oncologists use the information immediately to inform decision making. That’s why validation and quality control are so important.”
That said, there are no universal standards or guidelines as yet. “A new drug has to go through clinical trials, Food and Drug Administration approval and post-market follow up. But that’s not usually the case for molecular testing,” Duncavage acknowledges. “Most molecular testing is done in the setting of a laboratory-derived test, or LDT, where it's up to the individual lab to do quality control and ensure performance.”
CAP/CLIA laboratories, such as those at Washington University in St. Louis, must meet some additional requirements. However, NGS-based testing only became available in CAP/CLIA laboratories in 2012, and “the field has moved so quickly that we’re still developing the standards,” Duncavage says. “We at Washington University, along with other groups, are meeting with the FDA in an effort to define, for example, what kinds of experiments need to be done to demonstrate that an assay works, and what constitutes adequate validation. We've come up with some ideas that we've implemented for our own quality control, but what will eventually be required by all laboratories isn’t yet clear.”
Bioinformaticists, Molecular Pathologists, Where are You?
The biggest need right now—and this is similar for discovery NGS—is for people skilled in bioinformatics to make sense of the data and its implications, Duncavage emphasizes.
“We can implement robotics and other technologies to keep up with the sequencing demands. But our ability to understand the data has lagged far behind our ability to sequence. We need people to analyze and archive data, and we need people to develop new bioinformatics tools that will make those processes easier. This is a tremendous opportunity for professionals skilled in these areas because we and others are having a difficult time finding qualified people.” The unfulfilled need has led the Washington University in St. Louis to develop a bioinformatics training program as well as fellowships in molecular pathology to help move the field forward.
SLAS, too, understands this need and is offering a three-part webinar series titled Extracting Meaning From Your Data. Webinar presenters talk about experiment design, tools and technology and data analysis. Following live webinar presentation, all three sessions will be available through SLAS On-Demand. Live and On-Demand sessions are free for SLAS members.
Bioinformatics expertise could help solve some key problems that have emerged from NGS DNA sequencing, according to Duncavage. For example, “we don’t really know, even from a basic science standpoint, the full spectrum of structural variation in lung cancer, which we suspect is quite complex. One of the things we’re struggling with now is coming up with a reliable way to detect translocations, insertions, deletions and copy-number variation and to understand how common these are and whether they are clinically relevant.”
Despite the difficulties, the Washington University in St. Louis team is forging ahead. “Our gene panel is being used effectively here, and its use is reimbursable by insurers. Since we are a nonprofit academic laboratory, we make it available to any clinical service that wants it,” Duncavage says.
“In developing the panel, we had to take costs into account,” he adds. “The cost of data generation in large-scale genomic studies often exceeds $50,000 per patient. For our assay to be accessible to patients and insurers, we had to make sure we could keep the costs significantly lower—and we did. It’s still the early days, but I suspect that within the next year or so, this type of testing will be much more commonplace than it is now.”
SLAS2013 Diagnostics Track
Duncavage’s SLAS2013 presentation is part of the Diagnostics educational track, which features three sessions, each with multiple speakers: Advances in Near Patient Testing Devices, Diagnostics Tests and Personalized Medicine and Identification of Diagnostic Biomarkers with Novel Clinical Application.
Another presentation from the SLAS2013 educational track was featured previously in SLAS Electronic Laboratory Neighborhood. “From Laboratory to Clinic: Novel Skin Sampling Technique Simplifies Disease Detection,” reviewed genomics researcher William Wachsman’s work to diagnose melanoma by means of adhesive tape stripping, now known as epidermal genetic information retrieval.