It's getting there, but according to a study by researchers at Stanford, whole-genome sequencing (WGS) faces several hurdles before it's ready for the clinic.
As the cost of WGS edge toward the $1,000 mark, optimism has been growing that clinicians could routinely use WGS to diagnose people with unusual diseases, identify genes behind rare inherited disorders, and predict how patients might respond to treatments. But it seems that questions about accuracy may be holding back the widespread clinical use of WGS.
In a small pilot study, the Stanford team sequenced the full genomes of 12 adults. They found that current WSG technologies struggle to reliably and accurately read certain parts of the genome that contain important disease-related variations. To further hamper progress, interpreting the vast amounts of data generated by WGS still time-consuming and subjective. Results of the Stanford study were published in JAMA.
In spite of these drawbacks, Euan Ashley, one of the study's leaders, is quick to point out that there is still plenty of interest in the clinical use of WGS. "There's no question in our minds that whole-genome sequencing is transformative for medicine," he said. "But as real fans of the technology, we wanted to take a hard and realistic look at where we are today."
The Stanford team sequenced 12 healthy volunteers using two different technologies-one from Illumina and one from Complete Genomics. The platforms failed to capture 10 percent to 19 percent of genes associated with inherited diseases at enough depth to accurately identify all their nucleotides, which made it hard to link nucleotide variations to a person's health.
And the two platforms were not always consistent. They largely agreed over variations in single nucleotides, but rarely did so when it came to larger insertions or deletions (indels)-many of which are associated with disease risk. For these indels, the results from the two sequencing platforms agreed just one-third of the time. "These are solvable problems, but we have to be realistic that we haven't solved them yet," said Ashley.
The Stanford team used an algorithm of their own design to parse the data for medically relevant variants. They ended up calculating 89 to 125 per patient and manually reviewed each variant. They checked the literature and existing databases to determine whether these variants were likely to affect the risk of disease, assessed the odds as to whether the aberrations might change the structure of a protein, and more. They eventually winnowed the list down to around two to six clinically important variants per patient.
One female patient had a variant in the BRCA1 gene that meant she was very likely to develop breast and ovarian cancer, despite having no family history of either disease. A clinical genetic testing lab confirmed the result, and after discussions with her doctor, she opted to have her ovaries and Fallopian tubes removed, and to go through regular breast screening.
"Within this small group, there was a life-changing discovery," Ashley said. "While, given the size of this study, it's hard to extrapolate how often such cases would arise in a general population, it does speak to the promise of the technology."
On another positive note, the researchers found that the process was cheaper than expected. They estimated that total cost of WGS, including sequencing and interpretation of results, came in at $15,000 per patient. And based on the results, independent physicians recommended further tests at a cost of $351 to $776 per patient. "I'm often asked if healthcare costs will skyrocket if we do this for everyone," said Ashley. "Surely there will be hundreds and thousands of dollars spent on follow-up tests like MRIs and CT scans. But in this study, the follow-up costs were relatively modest."
However, interpreting the data was time-consuming and largely subjective. The genomics professionals who reviewed the variants spent about an hour on each, and often disagreed about their potential to cause disease.
"This report is sobering, but not at all surprising," Eric Topol from the Scripps Translational Science Institute in La Jolla, California, told The Scientist in an e-mail. "While WGS has indeed been useful in some individuals with a serious, undiagnosed disease, extrapolation to healthy volunteers is off the mark. All the accuracy problems become magnified."
"Not until we have millions of individuals with WGS, across various medical conditions and ancestries, will sequencing be adequately informative," Topol added. "We're at least a few years away for that to be the case."
David Dimmock from the Medical College of Wisconsin in Milwaukee had a different point of view. "The study is quite outdated," he said. "A lot of the limitations they bring up are things we were well aware of in 2011, and they're much less of a problem now because technology has progressed significantly. There are still holes, but we're in a very different place," he added.