NEW YORK (GenomeWeb) – Extracting pharmacogenetic variants from exome or genome sequencing data is feasible and generates more clinically useful information than current PGx chips, according to a recent study by researchers at the National Institutes of Health's National Human Genome Research Institute.
But although this capability may enhance the utility of clinical exome and genome tests, calling PGx variants from sequence data is still technically difficult and many clinicians resist using pharmacogenetic information when prescribing drugs.
Exome and genome tests have proven extremely useful for establishing molecular diagnoses in patients with rare genetic disorders where other diagnostic tests have come back negative, but their cost is still high and the tests generate a lot of information about the genome that is unrelated to the patient's condition.
"So to make it an efficient, useful, and cost-effective part of healthcare, what we have to do is figure out how to extract multiple uses of that one test in order to distribute its costs over as many benefits as we can," said Les Biesecker, chief of the medical genomics and metabolic genetics branch at NHGRI and the senior author of the study, which was published online in Genetics in Medicine last month.
One such additional use is to screen the sequence data for secondary findings, such as cancer risk variants in certain genes, that may benefit patients or their families, and many labs offering clinical exome tests already do so, often following the recommendations of the American College of Medical Genetics and Genomics.
Another potential use of exome data is to extract PGx variants from it, which might help doctors prescribe certain medications in the future. According to Sherri Bale, managing director of GeneDx, there is considerable interest in this among the clinical lab community. "We're definitely considering the importance of this and trying to figure out how to do it," Bale told GenomeWeb. This will involve both figuring out how to pull PGx variants from sequence data technically and how to report the information so it is most useful to clinicians.
Biesecker's team set out to study how much pharmacogenetic data an exome or genome would yield, and how such data would compare to data from a chip that is frequently used in pharmacogenetic testing.
For their study, the researchers selected 203 clinically relevant PGx variant positions, including 50 level 1A and 1B PGx variants from the Pharmacogenomics Knowledgebase (PharmGKB) and 154 variants from 40 gene-drug pairs with level A evidence from the Clinical Pharmacogenetics Implementation Consortium (CPIC).
They then looked for these variants in 973 exomes, five genomes, and five chip genotyping datasets — all from participants in the NIH's ClinSeq project. To generate the chip data, they used the Affymetrix Drug Metabolizing and Transporters (DMET) Plus array, which only targeted 60 of the 203 variants but is currently commonly used for PGx genotyping.
In five individuals for whom they had all three types of datasets available, the genome data identified 998 out of the 1,015 (five times 203) variants, the exome data found 849 of the variants, and the chip data just 295.
Overall, they identified 36 pharmacogenetic star allele variants with moderate to strong CPIC therapeutic recommendations in the 973 exomes, as well as 57 CNVs in the CYP2D6 gene.
The exome in particular turned out to be better at detecting important PGx variants than the team had expected. "We looked at the exome initially suspecting that it would be substantially inferior to the current clinical standard and were very surprised that it isn't," Biesecker said. Notably, the exome detected some PGx variants in noncoding regions that the test is not designed to target but picked up nevertheless.
The main reason the exome performed better than the chip is that the chip did not target most of the variants the researchers were looking for, suggesting that a more up-to-date chip would have performed better in the comparison. However, Biesecker said, "the problem with chips or panels of any type is, they are always going to be behind the current knowledge because the day after you design a chip, somebody publishes a study that shows a new pharmacogenetic variant is very important for some gene and some drug, and it's not on the chip."
Based on their comparison, for patients who already have their exome data available, the exome has the potential to provide more useful pharmacogenetic information than a commonly used chip that is run alongside the exome test, but extracting PGx information from the exome is not trivial.
"For a practicing clinician, today, the chip would win, because it's an established clinical test, and it can be ordered and provide an interpretation that's usable today," Biesecker said. "The problem is that there is no ready way, clinically, to do what we did, which is to turn exome data into pharmacogenetic alleles."
Researchers are already working on tools to pull pharmacogenetic information from sequence and array data, among them a collaboration between PharmGKB, CPIC, the Pharmacogenomics Research Network (PGRN), the Electronic Medical Records and Genomics (eMERGE) Network, and ClinGen to develop a tool called PharmCAT to translate genetic variants into pharmacogenetic star alleles and generate a report.
"It's a lot of manual labor to do those interpretations today on our data, so we will probably wait until the computational tool is available to semi-automatically extract those, and then we will probably run it on the entire [ClinSeq] cohort that we have already sequenced," Biesecker said.
However, even if the technical barriers are overcome, another challenge will be to convince clinicians to consider pharmacogenetic information when making prescriptions, he said. Up until now, this has often been impractical, costly, and time-consuming because it required ordering a separate PGx test prior to writing the prescription. "It does require a culture change, a social change, amongst prescribers to begin to accept this," he said. "We need to reduce the technical barrier to getting the information, and the time barrier and the cost barrier, because there is resistance in the clinical field to using pharmacogenetic data for prescribing. And I think our job is to make it as user-friendly as we possibly can in order to maximize people's willingness to try and use these data."