The results of a new study call into question the practice of sequencing patient's tumors without directly comparing the sequence data to the individual patient's healthy genome.
In work published in the April 15, 2015, issue of Science Translational Medicine, a team of researchers from Johns Hopkins University and Personal Genome Diagnostics LLC, showed that about two-thirds of mutations overall and a third of the mutations in so-called actionable genes – genes for which there are targeted therapies or candidates in clinical trials – were germline mutations that could be found not just in the tumor, but in the patients' normal cells as well.
For cancer patients, sequencing, both of targeted genomic regions and of the whole exome or genome, is not the standard of care yet.
But there are already tens of thousands of patients annually who are being sequenced, and Victor Velculescu told reporters at a press conference announcing the findings that "we expect this number to increase to potentially 1 million patients, representing all late-stage cancer patients."
And "although there are many factors that physicians take into consideration when deciding therapeutic options, such tests play an increasingly important role."
Velculescu is at Johns Hopkins University, and is the co-founder and chief scientific officer of Personal Genome Diagnostics.
Personal Genome Diagnostics was founded in 2010 and was the first company to do CLIA-certified exome sequencing. The company does gene panel sequencing and exome sequencing for clinical use as well as for biopharma companies,
"We'd always done tumor-normal [comparison] as our standard approach," chief commercial officer Antony Newton told BioWorld Today. "But we'd never gone back and proved that that was important."
The idea is to give patients and their doctors a detailed understanding of the molecular alterations of their tumors that will translate into an understanding of which targeted therapies are best suited to fighting those tumors.
Current diagnostic tests most often zero in on specific mutations in specific genes to see whether a tumor might be susceptible to a targeted agent.
Next-generation sequencing, Velculescu said, "is looking at many more genes, and looking at all the bases in those genes."
But it does such sequencing only for the patients' tumors. And the results patients get can be a case study in too much data, not enough information.
"Analyzing only tumor data may not be sufficiently accurate," Velculescu said, because purely from the sequence data, there is no way to tell whether a mutation in a cancer gene is actually driving cell growth, or is functionally silent.
If a mutation is found in both the normal and the tumor genome, it is a strong indicator that it is not a cancer driver.
In their studies, Velculescu and his colleagues sequenced the tumor and regular genomes of roughly 800 patients.
About three-quarters of those patients had at least one mutation in an actionable gene. But the team found that even after filtering out the best-known germline alterations, two-thirds of the remaining mutations and one-third of the mutations in actionable genes, were not tumor-specific, but also found in the germline.
There are variants, even within genes that are bonafide drug targets under certain circumstances, that have only minor effects, Velculescu explained.
"But they may occur in a particular gene or a region of the gene that looks very similar to an actionable mutation. And so there are mutations that may be masking themselves as actionable mutations," he explained
Because each patient on average had more than one alteration, the average chances of having a non-actionable mutation in an actionable gene were even higher.
"False-positive changes affected roughly one in every two patients analyzed," Velculescu said.
The team's conclusion is that to truly benefit from next-generation sequencing, sequencing of the healthy genome as well as the tumor genome will need to become routine.
"Precision medicine depends on precision genomics," Velculescu said.
Such extra sequencing is easier said than done, of course.
There are logistical concerns of getting a tissue sample for comparison sequencing, though because there is no gene expression profiling involved, the comparison sample could come from an easily accessible source such as blood or saliva. Personal Genome Diagnostics' Newton said that "the majority" of the patients his company sequences "have a tumor and a normal [sample] come in."
Of more serious concern is the loss of patient privacy that comes with whole-genome sequencing. (See BioWorld Today, June 4, 2014.)
Velculescu said that such concerns could be addressed by using bioinformatics methods to make only the subtracted data on tumor-specific alterations available to clinicians.
Finally, there is the question of who will pay for the extra sequencing.
Velculescu said that currently, insurance does not fully cover the cost of such additional sequencing.
There is an ongoing discussion about those costs.
The high false-positive rates that the team reported certainly suggest that the cost of a matched tumor-normal test, which is in the range of $4,000-$5,000, is probably a good investment.
"It's a lot of money if you're paying out of pocket," Newton said. "But it's not that much compared to the cost of a targeted therapy."