Skip to main content
Premium Trial:

Request an Annual Quote

Technology, Sample Challenges Continue for Clinical NGS, but Becoming More Solvable


NEW YORK (GenomeWeb) – Over the years, laboratories looking to bring next-generation sequencing into the clinic have detailed numerous challenges — from accuracy of the sequencing, sample prep, data interpretation, to turnaround time, among many concerns.

Researchers who spoke during a panel discussion at the Festival of Genomics conference in San Mateo, California this week, said that although these challenges continue to persist, they are more or less solvable as long as you know what the problem is and the correct technology to use to solve it.

Nonetheless, the researchers said that for the vast majority of cases, whole-genome sequencing is not yet clinically viable.

One major issue, according to Jennifer Morrissette, clinical director of the Center for Personalized Diagnostics at the University of Pennsylvania, is working with formalin-fixed paraffin-embedded tissue.

She said when the lab first began developing a clinical sequencing panel for samples from FFPE tissue, they initially found that between 10 percent and 12 percent of all samples could not be run on the panel, either because there was not enough DNA or the quality was just so bad. She said the lab developed a test specifically to try and salvage these samples.

The assay, dubbed the Penn Precision Panel, requires just 250 picograms of starting material and looks at 20 genes, Morrissette said, as opposed to the more commonly used 47-gene solid tumor panel. It's for "really bad samples," she added. However, the incorporation of that panel reduced the percentage of samples that couldn't be tested at all to 1 percent.

In addition, said Morrissette, sample quality can vary depending on how the sample is fixed in the first place. While the general rule of thumb has often been that the longer a sample has been fixed, the more degraded, she said it's actually more dependent on how long the FFPE has been cut. "If it's been cut and sitting on a bench, that matters more than how long it's been sitting in the block," she said.

Robert Sebra, director of technology development and assistant professor at the Icahn School of Medicine at Mt. Sinai, agreed. He said sample degradation can occur before the sample has even been put in FFPE. A lot depends on the practice of the surgeon, he said, and improvements can be seen just by talking to the clinicians and surgeons.

In some cases, Sebra added, studies can be amended such that a small portion of the sample is flash frozen and the rest is fixed. He noted one example where Mt. Sinai was able to change a protocol to enable single-cell transcriptome sequencing.

However, implementing this dual process — freezing and fixing — into everyday practice would be challenging. There would then need to be liquid nitrogen in the operating room.

Aside from sample quality, different clinical questions are better answered by different technologies. For instance, Sebra's team at Mt. Sinai created a whole-genome human de novo assembly using technologies from BioNano Genomics, Pacific Biosciences, and Illumina.

"In an ideal world, we'd do this for every genome," Sebra said, but realistically that's not possible, especially in a clinical setting. The long-range information provided by BioNano Genomics and Pacific Biosciences were critical for understanding structural variants, he said, which can't always be obtained with shorter-read sequencing technology.

Another instance where Sebra's team has found long reads to be especially useful is for diagnosing diseases that are caused by mutations in genes that are especially large or have pseudogenes. For instance, Gaucher's disease is often caused by mutations to the gene, GBA. Over 200 mutations in the gene have been found to cause the disease, making clinical presentations very diverse. As such, the disease is hard to diagnose by clinical presentation alone.

Sebra said that his lab at Mt. Sinai is working to design a sequencing-based test using a 6.4 kb long-range amplicon that it will sequence on the PacBio RSII. But, the presence of the pseudogene makes the design of the amplicon tricky. "You have to get the flanks situated correctly to make sure you're in the gene and not the pseudogene," he said.

Morrissette said her lab has also had issues in designing gene panels for genes that aren't amenable to the technology the lab was using.

For example, when designing an NGS panel for hematological malignancies, Morrissette said that her team noticed that the panel was not good at identifying internal tandem duplications in the FLT3 gene, which are critical biomarkers for predicting poor prognosis in patients with acute myeloma leukemia.

She said the researchers went back through the sequence data to try to understand why those alterations were not being detected. From samples that the lab knew were FLT3-ITD positive, they noticed that after a quality filtering step, the number of reads were reduced dramatically. It turned out that in the algorithm, there was a mapping step that would throw out reads that did not align to the reference. Because the internal tandem duplications in the FLT3 gene caused a mismatch, reads with that alteration were being filtered out. To fix the issues, Morrissette said the researchers designed a program called Garbage Picker, which went back through the trashed reads and salvaged the ones with the alteration.

David Smith, professor of laboratory medicine and pathology at the Mayo Clinic, noted that there is still a big gap between NGS in the research world and NGS in the clinical world. Although sequencing prices have come down dramatically, he said, those prices don't take into consideration extra costs that are specific for the clinical realm, like data storage, he said.

Currently, the Mayo Clinic offers clinical NGS-based gene panels, he said, but those panels are all very small, in part because the center is required to store patient data for 20 years. Even gene panels that evaluate a few hundred genes "are not a viable business model" when storage costs are taken into consideration, he said. And a clinical whole genome would cost closer to $20,000 than $1,000, he said, when all costs are taken into consideration.

Morrissette, too, said the scope of the NGS panels offered at Penn's Center for Personalized Diagnostics are driven by cost and the amount of information the clinicians want.

Clinicians "want an answer to a specific question," she said. They want to know if there is a diagnostic, prognostic, or therapeutic target. "They don't want any additional information about their cancer patients other than something they can act on."