Whole Genome Sequencing Provides Fast, Accurate Diagnoses in the NICU

By Allison Proffitt  

October 3, 2012 | Using two new algorithms and Illumina’s newest sequencer, doctors at Children’s Mercy Hospital in Kansas City, Missouri, have used whole genome sequencing to rapidly diagnose genetic diseases in acutely ill newborns. The results were published today in Science Translational Medicine (3 October 2012, Sci. Transl. Med. 4, 154ra134 (2012)).  

One of every 20 babies born is admitted to the neonatal intensive care unit at birth, said Stephen Kingsmore. At Children’s Mercy Hospital in Kansas City, where Kingsmore is director of the Center for Pediatric Genomic Medicine, the leading cause of these admissions is an illness “that’s likely to be genetic.” But with over 3,500 monogenic diseases known, the identity of these illnesses is distressingly elusive.  

Kingsmore’s team from the Center for Pediatric Genomic Medicine including Neil Miller (director of informatics) and Carol Saunders (clinical laboratory director) worked closely with Kevin Hall, director of systems integration at Illumina in Essex, United Kingdom to develop a whole genome sequencing approach to diagnosing these diseases.  

Kingsmore is no stranger to baffling genetic diseases. In 2010, while he was the director of the National Center for Genome Resources (NCGR) in Santa Fe, New Mexico, Kingsmore worked closely with Craig Benson to develop the Beyond Batten carrier screening panel (see, A Legacy for and Beyond Batten Disease). But with acutely ill babies, timing—not just accuracy—is paramount.   

The proof of principle solution that Children’s Mercy Hospital tested was a combination of new hardware and software tools, all optimized for speed.  

Whole genome sequencing for the test was done on an Illumina HiSeq 2500 prototype. The new sequencer model is built on the same core architecture as the HiSeq 2000 and features advances in cycle time reduction and onboard cluster generation originally developed for the MiSeq. The HiSeq 2500 is capable of generating 120 Gb in 27 hours in rapid run mode.  

Two informatics tools were developed to process the sequencing data as efficiently as possible. SSAGA, symptom- and sign-assisted genome analysis, matches 227 clinical terms from nine symptom categories to 591 well-established, recessive genetic diseases with pediatric presentations and then helps prioritize the clinical information and interpret WGS results in light of that data. Clinicians—even those not familiar with all of the genetic disease options—choose the clinical terms that reflect the patient’s symptoms, and SSAGA identifies likely genetic culprits. When tested against over 500 past patients, SSAGA identified the correct disease and suggested the affected gene in 99.3% of cases.  

The next question is, “What effect might each of the variants that we detected in each patient have?” explained Neil Miller, director of informatics at the Center for Pediatric Genomic Medicine at the Children’s Mercy Hospital. Another new software pipeline tackles this problem. RUNES, rapid understanding of nucleotide-variant effect software, characterizes each of the four million variants found in a person’s genome, identifying known disease-causing variants and using predictive tools to estimate variant consequence. RUNES then assigns a score that estimates how likely the variant is to be disease-causing. 

Using SSAGA and RUNES, much of the onerous characterization of genome variation and interpretation was bypassed by restricting and prioritizing variants with respect to allele frequency, likelihood of a functional consequence, and relevance to the prompting illness, said study authors in the paper. 

“We use SAGA to determine what genes to look at, and then use RUNES to determine which genes look like they’re being disrupted, or in a disease state, and together we’re able to narrow down that list of four million variants to a handful which can then be reviewed by a clinician,” said Miller. “Software is great at doing large-scale repetitive analysis, but ultimately you need expert medical review to look at the results and make the diagnosis.” 

Because the technique looks at only the specific genes associated with phenotype that appear to be disrupted, it is appropriate for prompt disease diagnosis rather than carrier testing or newborn screening.  

Time Sensitive  

The study authors report that the process can be completed in 50 hours from initial blood draw to delivering a verbal, interim diagnosis to the ordering physician. The HiSeq 2500 can generate 120 Gigabases in 27 hours and the Illumina CASAVA tool was used for alignment. Sample preparation took 4.5 hours instead of the standard 16 hours so that the total “hands-on” time was only about 5 hours, said Illumina’s Hall. SSAGA-delimited variant analysis and interpretation takes only an hour; RUNES analysis runs in about 2.5 hours per sample.  

But there are caveats. The 50-hour time frame is “elapsed time”, Kingsmore said. Because Children’s Mercy sent the blood samples to Essex, U.K. for sequencing on the HiSeq 2500 prototype, there was transportation time to contend with. The “50 hours” number counts time spent actually collecting sample and doing the sequencing and analysis and not the transit time to and from Essex. Once Children’s Mercy has its own HiSeq 2500, Kingsmore expects the two-day target to be easily met and believes it could even be lowered to 36 hours.  

The test results are also subject to confirmatory testing, which can take two to three days. Because of CLIA guidelines on medical tests, the results of the test are only available to the ordering physician as an interim report. In order to share the findings with the families and enter the data into the medical record, the mutation-causing regions need to be re-sequenced with Sanger sequencing in a CLIA-certified laboratory.  

Kingsmore believes this won’t always be the case, however. “For about the last 18 months, our team here at Children’s Mercy has been working with this type of next-generation sequencing to have a specific set of diseases meet those guidelines,” he said. “Over the next month, we should be able to make that a routine test and report the results in the medical record.” 

Testing in Light of Poor Prognosis  

The proof of principle study was based on seven cases of monogenic disease. Two of the cases were ones in which the approach was used retrospectively to accurately diagnose children who had died. Five were undiagnosed cases, and of those, four were diagnosed.  

Sadly, none of the children were saved thanks to their quick diagnoses. Sequencing revealed fatal mutations. But the tests do still have great value for the patients.  

“Having a definitive diagnosis is useful for families, even in cases like these where there’s no effective treatment,” said Carol Saunders, lead author on the paper and clinical lab director at Children’s Mercy. “Knowing the marker or defect may provide some information regarding the prognosis so the family knows what to expect. Importantly, it also allows them to have accurate genetic counseling regarding their risk to have another affected baby, and to make informed decisions about their reproductive future.” 

Kingsmore agrees. “Of the 3,500 conditions or genetic diseases that we know the gene for, there are roughly 500 for which some type of treatment is available. Now not all of those treatments are curative. Many of them are partially effective... But I think Carol made a really important point, that in the remaining cases where the baby will die, and where the prognosis is dire, this still can have tremendous benefits.” 

The test is expensive; Kingsmore estimates about $13,500. But he has already seen acceptance by neonatologists at Children’s Mercy and believes it will be available by the end of the year. Miller, director of informatics, says SSAGA is being expanded to cover all 3,500 know disease genes.  

“We think also that we should be able to offer this as a service to physicians around the country, who are able to ship us a DNA sample,” once Children’s Mercy has its own HiSeq 2500 instrument, said Kingsmore. “It will take a little bit longer than two days, but in that manner I think we should be able to offer this pretty generally sometime early next year.”