Researchers within the Pharmacogenomics Research Network and the Electronic Medical Records and Genomics, or eMERGE, consortium have teamed up to conduct a study that will integrate genomic data gathered from a sequencing-based pharmacogenomics panel with the electronic medical records of as many as 10,000 patients.
At last week's American Society of Human Genetics annual meeting in San Francisco, Adam Gordon, a graduate student in Debbie Nickerson's laboratory at the University of Washington, discussed the so-called PGRN-seq panel, which targets 84 genes associated with drug response.
The team has tested the panel on 32 trios from the 1000 Genomes Project as well as 94 patient samples with extensive genotype and phenotype data, and plans to launch a pilot study with the eEMERGE network that will test the panel in 7,000 to 10,000 patients across all nine eMERGE sites. The project intends to integrate the clinically actionable data into patients' medical records and to deposit all of the data into a centralized database for further study.
There are seven adult eMERGE sites: Vanderbilt University, Northwestern University, Mount Sinai School of Medicine, the Mayo Clinic, the Marshfield Clinic, the Group Health Cooperative with University of Washington, and Geisinger Health System. Additionally, there are two pediatric eMERGE sites: Children's Hospital of Pennsylvania and a dual site shared between the Cincinnati Children's Medical Center and Boston Children's Hospital.
The PGx panel encompasses three different types of genes — drug targets, those involved in drug transport, and those that impact drug metabolism. The group sought to choose not only well-validated genes, but also genes associated with commonly prescribed drugs, such as warfarin, clopidogrel, and statins.
Using Nimblegen in-solution enrichment technology and Illumina sequencing, the panel captures all of the exons of the 84 genes, the untranslated regions, and 2 kilobase pairs upstream of the exons.
Twenty four samples are multiplexed per sequencing lane, which enables an average coverage per sample of around 500-fold, Gordon said. He added that the researchers also tested 96-plexing, but that resulted in lower coverage. High coverage is necessary because "we want to discover rare variants and know that they are rare," he said. Additionally, in the future, the team would like to do copy number analysis.
The team tested the panel on 96 samples (32 trios) from the 1000 Genomes Project and 94 patient samples and found that they were able to identify rare variants. The panel also yielded results that were concordant with the whole-genome sequence data.
They also found around 1,300 to 1,500 SNVs per sample, about 30 to 50 of which were novel. Additionally, when doing functional analyses of the novel variants, they estimated that 9 percent of them were putatively functional.
For example, Gordon highlighted a novel nonsense variant in the RYR1 gene, which was present in one of the fathers from the 1000 Genomes set. The variant was initially missed because sequencing was done at low coverage, but it was found in the deeper sequenced panel.
The gene has recently come under scrutiny for its role in malignant hyperthermia, a rare but serious and sometimes fatal side effect to anesthesia. "A lot of disruptive mutations have been found in [RYR1] in patients with malignant hyperthermia," Gordon said. "So this could potentially be serious, but it's in a heterozygous state, so it's unclear whether this would have an effect."
Additionally, the team identified rare missense variants in AHR, a gene implicated in drug metabolism pathways associated with caffeine, nicotine, albuterol, and other xenobiotic compounds, Gordon said.
When implementing the panel in the clinic, said Gordon, the challenge will be to figure out which variants to include in the patient report and which to deposit into the centralized research database.
In the case of the RYR1 variant, he said that would likely be included in the report, while the AHR variants would likely go into the database. However, he noted, those decisions would ultimately be made by the individual institutions implementing the panel.
Gordon said that he is now working with the other sequencing centers within the Pharmacogenomics Research Network — Washington University and Baylor University — to optimize the sequencing protocol and to make sure that all centers obtain consistent results.
Additionally, the nine institutions within the eMERGE Network are also working out the logistics involved in the initial pilot.
Dan Roden, assistant vice chancellor for personalized medicine at Vanderbilt University, which is the coordinating center for the eMERGE sites, told Clinical Sequencing News that the study has two primary goals: First, to figure out whether genomic information can be used to predict patients' drug response and, second, to figure out how to incorporate genomic information with electronic medical records to impact patient care.
While the panel screens 84 genes, Roden said that only variants that have been well validated and proven to be clinically actionable will be reported in the patient's record. However, all of the data will be deposited into the centralized research database. Each individual site is now coming up with its own list of which variants to include in EMRs.
For instance, he said, a well-validated variant in CYP2C19 that predicts response to clopidogrel is on the list for five out of seven of the adult sites. Two genes associated with warfarin response, VKORC1 and CYP2C9, are also on the list for five out of seven sites.
And the SLCO1B1 gene, variants to which can affect adverse events associated with simvastatin, is currently on the list for six sites and the seventh site is still considering it.
Clopidogrel, warfarin, and simvastatin are among the drugs to be included, not only because there is evidence about how a person's genetics can impact their response, but also because they're commonly used drugs, Roden said. "Other drug-gene pairs have as good as or better evidence but are not as widely used," he said. And because the group is taking the pre-emptive strategy of genotyping patients before they are prescribed a drug, "if you target these commonly used drugs, it's more likely that you're going to end up with people that actually use the drug."
Each site will screen around 1,000 patients that have been identified as being the most likely to be prescribed a drug for which the panel screens.
The goal, said Roden, is to then implement a system so that when a physician logs into the patient EMR to prescribe a drug, if the patient has a genomic variant impacting drug response or indicating an adverse reaction, the prescription will trigger an alert to the physician.
Roden called the project "preemptive genotyping." Rather than waiting until a physician writes a prescription to do a genetic test, the goal is to identify patients likely to receive to a drug for which there are known genetic variants impacting response, and genotype them ahead of time, so the information will be readily available when needed.
The project is still in its early stages, Roden said. The sites are still working out the list of variants that will be included in the report and developing the algorithms that will identify the patients to screen and the clinical decision infrastructure.
However, he said, Vanderbilt has already implemented a similar program, dubbed PREDICT, for Pharmacogenomic Resource for Enhanced Decisions in Care and Treatment, which is based on microarray technology. Components of that system will likely be incorporated into the sequencing-based program, he said.
Other groups have also found value in using sequencing to predict drug response or find additional variants related to drug response. A group from the Medical College of Wisconsin is working on a 35-gene pharmacogenomic panel to be run on the Ion Torrent PGM and researchers at Scripps Translational Science Institute are using exome sequencing to identify markers predictive of clopidogrel response (CSN 1/11/2012 and CSN 3/21/2012).