Skip to main content
Premium Trial:

Request an Annual Quote

Whole-Genome Sequencing Has Better Dx Rate Than Exome at Similar Cost, Australian Team Finds

NEW YORK (GenomeWeb) – Researchers at the Garvan Institute of Medical Research in Australia said last week that they can diagnose rare, monogenic diseases more than half the time using whole-genome sequencing.

According to Mark Cowley, who presented the work at the Biology of Genomes meeting at Cold Spring Harbor Laboratory last week, this diagnostic rate is much better that the 25 percent rate typically observed with whole-exome sequencing.

His lab is currently undergoing an assessment by the Australian National Association of Testing Authorities — similar to CLIA certification in the US — to offer clinical whole-genome sequencing to diagnose patients with rare genetic disorders. It has already used this approach to diagnose patients with autosomal dominant polycystic kidney disease, hereditary spastic paraplegia, and other disorders.

"Ultimately, the goal we are trying to achieve here is to start off in a clinical environment with patients and their families seeing clinical geneticists, being counseled for the test, going through sequencing, bioinformatics, and clinical interpretation [to produce] a clinical report," Cowley said. "Hopefully, this clinical report returns a positive diagnosis for the patient … and improves the clinical management of that patient."

While whole-genome sequencing picks up coding and structural variation that exome sequencing may miss, clinicians and researchers often turn to exome sequencing because of the reduced cost. In the US, Cowley noted, a genome sequence could cost three times as much as an exome sequence.

However, Cowley said that because of Illumina's pricing strategy, the difference in cost between whole-genome sequencing and whole-exome sequencing is smaller in his lab than in others.

Institutions that run a lot of exomes, he noted in an email to GenomeWeb, have stronger buying power and, because of that, their reagents, and thus exomes, are cheaper than at other, lower-bulk institutions.

But since the Kinghorn Centre for Clinical Genomics at Garvan is one of the first three Illumina HiSeq X Ten sites, and all the HiSeq X Ten labs pay the same price — the price reserved for large labs — for their reagents, they get a break on price there. Added to that is the fact that his lab charges patients less overhead than most US labs do.

That, he said at the meeting, boils down to a clinical genome in his lab costing only 1.1 times as much as a clinical exome.

With their HiSeq X Ten, Cowley and his lab have run nearly 5,000 genomes and worked out some of the growing pains. Currently, he said, they have a 30X minimum depth of coverage, which increases to 40X coverage for coding regions. That, he added, gives them a lot of power to detect variants.

Sequence data they generate flows into a bioinformatics pipeline they developed based on BWA and GATK best practices using joint calling, which they've since moved to the cloud. While the structural variant and copy number variant pipeline isn't yet clinically validated, he noted that the small variant pipeline is. The copy number pipeline relies on CNVNator and structural variant pipeline on Lumpy.

The variants are then fed into an in-house variant filtration platform called SEAVE that helps clinicians to quickly diagnose patients, Cowley said.

To test the pipeline, Cowley and his colleagues turned to the NA12878 cell line. The sensitivity, he said, appears to be quite high — in the 99 percent range for SNPs and around 97 percent for indels. Most false positives, he added, appeared to disappear at about 15X depth of coverage.           

Using this approach, they were able to diagnose some 53.5 percent of a cohort of 308 patients.

For instance, despite its reputation as a disorder that is tricky to diagnose because of a high number of pseudogenes that are similar to the causal PKD1 gene, they obtained a 86 percent diagnostic yield for autosomal dominant polycystic kidney disease in a cohort of 28 patients. Most of the mutations they uncovered were loss-of-function mutations, and all of them were confirmed using Sanger sequencing.

They'd previously attempted to use whole-exome sequencing to diagnose autosomal dominant polycystic kidney disease in a cohort of 14 patients, but Cowley said they were disappointed with the 50 percent diagnostic rate they achieved. A few of these exome patients also underwent whole-genome sequencing, which picked up a splice site variant and a small deletion that exome sequencing missed. Cowley attributed that oversight to shorter reads and, possibly, to capture bias.

Similarly, in hereditary spastic paraplegia, Cowley and his team were able to pinpoint the disease cause in four out of five consanguineous families. One family, he noted, had a mutation in the PEX16 gene, which isn't a traditional hereditary spastic paraplegia gene and likely wouldn't have been caught by a targeted screen.

The approach can also be used to end patients' diagnostic odysseys, Cowley said. In another example, he and his team sequenced the genomes of two sisters with severe epilepsy and myoclonic-atonic seizures and their parents to find that both girls harbored an insertion that's not present in blood samples from either of their parents, suggesting mosaicism in one parent. Neither an epilepsy panel nor exome sequencing caught the insertion, he noted.

To speed diagnoses, his group has developed a patient database and collaborative diagnosis platform called Patient Archive.  According to Cowley, clinical notes from a geneticist can be pasted in and the key Human Phenotype Ontology (HPO) terms are pulled out of the free text.

This can be used, he added, to power gene prioritization algorithms and for sharing of de-identified patient data for patient matchmaking. He noted that it is to be rolled out through the Australian Genomics Health Alliance and the Undiagnosed Disease Networks in Japan and Australia.