As a first step toward implementing clinical whole-genome sequencing in a neonatal intensive care unit, researchers from Children's Mercy Hospital have published a proof-of-principle study demonstrating that their protocol can diagnose genetic disease in newborns.
Moving forward, the team plans to begin offering the so-called STAT-Seq test, which runs on Illumina's HiSeq 2500, on a research basis from its hospital until it receives CLIA certification.
Additionally, the hospital plans to launch a CLIA-validated test on the MiSeq later this month that sequences more than 500 genes associated with around 600 inherited diseases (CSN 8/9/2011). The team had originally developed the test on the HiSeq 2000, but is launching it as a clinical test on the MiSeq through a collaboration with Illumina on the company's TruSight Inherited Disease panel (CSN 9/12/2012). The test will have an out-of-pocket cost of $1,250.
The STAT-Seq proof-of-principle study, which was published online today in Science Translational Medicine, demonstrated that sequencing on the HiSeq 2500 machine, which can sequence a whole genome in around 26 hours, combined with the hospital's internally developed automated analysis pipeline could provide an answer in around 50 hours for an estimated cost of $13,500.
"We can now consider whole-genome sequencing to be relevant for hospital medicine," Stephen Kingsmore, director of Children's Mercy Hospital's Center for Pediatric Genomic Medicine, said in a conference call discussing the study.
"We think this is going to transform the world of neonatology by allowing neonatologists to practice medicine that's influenced by genomes," he added.
In the study, the actual time to result was around one week because the sequencing was done at Illumina's UK facility, so data had to be shipped between the Kansas City, Mo.-based hospital and the UK.
But turnaround time will be 50 hours when the hospital is able to run the test from its own facilities. Furthermore, Kingsmore said that by the end of the year the team will be able to reduce turnaround time to less than 36 hours through improvements to alignment and variant calling.
This summer, Kingsmore presented initial data from some early cases of the STAT-Seq protocol using the HiSeq 2500 to diagnose infants admitted to Children's Mercy Hospital's neonatal intensive care unit with suspected genetic disease (CSN 7/18/2012).
Kingsmore told CSN this week that STAT-Seq would initially be offered to patients in the NICU at Children's Mercy Hospital because for those patients there is an "intense urgency." Additionally, while an economic analysis has not yet been done on the protocol, beds in the NICU can cost upwards of $8,000 a day, and a series of single-gene tests will quickly add several thousands of dollars.
While $13,500 is not cheap, "a few days saved or more effective treatments" could lead to a greater financial gain, though he noted that such an economic analysis has not yet been done.
Other newborns or pediatric patients with suspected genetic disease for whom urgency is not as critical could be offered the hospital's inherited disease panel test, Kingsmore said. Despite the fact that this test is being run on the MiSeq and only looks at around 500 genes rather than the whole genome, the STAT-Seq protocol is quicker simply because it doesn't require an enrichment step, Kingsmore said.
Automated Analysis
A key to making STAT-Seq suitable for the clinic was in the team's analysis pipeline. Whole-genome interpretation is frequently cited as a major barrier toward implementing sequencing in the clinic.
Two software tools that the CMH team developed automate much of the analysis, however.
One tool, dubbed SSAGA, for symptom- and sign-assisted genome analysis, provides a framework for physicians to input a patient's clinical phenotype, and then it spits out a candidate list of disease genes associated with that phenotype, thus narrowing analysis of the entire genome.
Currently, the program has a menu of 227 clinical terms associated with 591 genetic diseases, but it is being expanded to include all 3,500 genes that are known to be causative of genetic disease.
The algorithm also lowers the probability of finding disease-related mutations that are unrelated to the patient's condition, since initial analysis is limited.
The researchers tested the algorithm on 553 children who had already received a molecular diagnosis, and found it had a sensitivity of 99.3 percent and nominated an average of 194 genes per patient.
SSAGA "lets a physician order a complex test in a simple manner," explained Neil Miller, the director of informatics at Children's Mercy Hospital's Center for Pediatric Genomic Medicine, in the conference call. The physician "clicks buttons for symptoms" and SSAGA "maps the symptoms to genes."
Once the initial gene list is narrowed down, variants in those genes are analyzed using a second software tool called Runes. Runes determines which variants are known to be disease variants and runs predictions about the functional consequences of unknown variants. For unknown variants, the program assigns a score to estimate the likelihood that the variant is disease causing.
"That narrows it down to a handful that can be manually reviewed," Miller said.
Proof-of-Principle
In the Science Translational Medicine study, the STAT-Seq protocol correctly diagnosed two retrospective samples, one of which had Tay Sachs disease and the other which had Menkes disease. It also provided a diagnosis for four out of five of the undiagnosed newborns.
The team tested the protocol in a research setting on five newborn patients that had been admitted to the neonatal intensive care unit at Children's Mercy Hospital. Because it was done in a research setting, confirmatory testing had to be done, which took at least an additional four days before results could be returned to the parents.
In one patient with severe epilepsy, a molecular basis was identified within just one hour of data analysis, despite the fact that the variant in the disease gene had only recently been reported in two unrelated Amish infants and was not yet in the reference databases.
The patient ultimately passed away after five weeks of failed treatment when the parents chose to end ventilatory support. While this decision was made before the parents received the diagnosis, Carol Saunders, the clinical lab director at CMH, said the diagnosis was still useful.
"As a result of this project, the family has received accurate genetic counseling, and can now pursue carrier testing for at-risk individuals," she said during the conference call.
In a patient with a severe skin disease, the team identified a novel de novo mutation in a gene known to cause lethal skin disease. However, because it is a novel gene, "we need more evidence," said Saunders, adding that the team is now collaborating with another group to determine if the variant was in fact causative. This patient also passed away.
"This was the first baby for these parents, and they are interested in trying for a healthy baby, but understandably would like to know their risks for this to happen again," Saunders said.
In a third family, the researchers think that they may have identified an entirely new disease gene. In this case, the parents had two affected babies with heart defects and a condition in which some of the internal organs are on the wrong side of the body.
The initial search did not turn up any known disease genes. But because the unaffected parents had two affected children, the researchers also sequenced the parents and looked for variants for which the parents were heterozygous but that were shared by the siblings and identified two variants in the gene BCL9L. While it had not previously been reported as a disease gene in humans, previous significant animal modeling further supported the identification of a new disease gene.
Additionally, the parents now know that they are each carriers, which "allows them to have accurate genetic counseling regarding their risk to have another affected baby, and to make informed decisions about their reproductive future," Saunders said.
In the fifth case, no molecular diagnosis was made. The patient died after five days and doctors suspected a mitochondrial disorder. In this case, Saunders said that the causative mutation likely falls outside of their detection abilities.
"There are many important variants that lie outside the regions that we typically know to analyze, and some types of variants aren't captured very well by this technology yet," she said, citing large copy number variants and repeat expansions as examples.
Since the study, the team has analyzed another two patients. In one case a molecular diagnosis has led to the parents pursuing pre-implantation genetic screening in order to have a healthy child, and in the second case, the researchers believe they may have identified another novel disease gene.
CLIA Certification
Kingsmore told CSN that ultimately the goal is to be able clinically validate the STAT-Seq test, but he anticipated that getting CLIA certification would take some time. For instance, he said, the CLIA certification process for the targeted panel test has taken around 18 months.
"Having done it once, it may be faster, but we just don't know," he said. Until the test is CLIA compliant, the hospital will offer it on a research basis, with initial results being reported to the physician verbally and then included in the official report after confirmation with Sanger sequencing or another clinically validated method.