Researchers at the Mayo Clinic are forging ahead with a pilot project that will integrate sequencing-based pharmacogenomics data into the electronic medical records of more than 1,000 patients and allow doctors to consider the information when they prescribe certain drugs.
The project is a collaboration between the Pharmacogenomics Research Network (PGRN), the Electronic Medical Records and Genomics Network (eMERGE), and the Mayo Clinic Center for Individualized Medicine, and is one of nine such pilots under the PGRN-eMERGE partnership.
All pilot studies use an 84-gene panel, called PGRN-seq, that contains genes relevant for drug response and side effects. The genes are enriched using NimbleGen in-solution technology and analyzed on Illumina sequencers.
The Mayo Clinic is one of only two eMERGE sites – the other one is Mount Sinai School of Medicine – that are implementing the PGRN-seq test in a CLIA-approved and CAP-certified clinical environment. Sequencing will be conducted at the Mayo Clinic Clinical Genome Sequencing Laboratory and the Personalized Genomics Laboratory.
According to Suzette Bielinski, the study leader and an assistant professor in the division of epidemiology at the Mayo Clinic, setting PGRN-seq up as a clinical test will allow them to include more variants in the EMR than labs that do the sequencing in a research environment and then have to confirm variants by a clinical genotyping test.
And, she said, while targeted genotyping tests, like microarrays, are a cost-effective way to collect SNP variants, they often do not capture other types of variants, and clinically important variants cannot be added at a later point, whereas sequence data includes all those variants.
The study protocol used by the Mayo researchers, called Right Drug, Right Dose, Right Time – Using Genomic Data to Individualize Treatment, or RIGHT, protocol was recently described in the Mayo Clinic Proceedings.
According to the authors, the project "provides an opportunity to begin to evaluate the impact of preemptive sequencing and [electronic medical record]-driven genome-guided therapy" and "improve understanding and implementation of genomic data in clinical practice."
For the project, the researchers recruited 1,013 patients from the Mayo Clinic Biobank cohort that were at increased risk of needing statin therapy within the next three years.
Over the next few months, they plan to run the PGRN-seq test on these samples, with the goal to complete all sequencing by June.
The PGRN-seq assay is currently undergoing verification testing with 96 control samples from the Coriell Institute in order to make sure it meets the standards of clinical testing.
"Trying to implement a test that was developed as a research tool into the clinical lab is challenging," Bielinski said. For example, according to current standards, the sequence coverage must be at least 100x, a goal that is not met across the entire target DNA with the in-solution capture of the research test. "We've been struggling to optimize areas that we really want to perform well to meet the clinical standard," she said.
Also, the CYP2D6 gene, which is involved in the metabolism of many important drugs, currently cannot be sequenced reliably by any of the existing short-read next-gen sequencing platforms, so the researchers decided to genotype it using an existing PCR- based test from Luminex that the Mayo Clinic has used for several years. CYP2D6 testing has already been completed for all patients in the study and the results are in their EMRs, Bielinski said.
Once the sequence data is available, the researchers initially plan to include in the EMRs only four variants that have been linked to the drugs warfarin, clopidogrel, and simvastatin. That number is expected to grow over time, and an expert committee at the Mayo Clinic continues to review other variants that are clinically actionable for inclusion.
When a doctor prescribes one of the three drugs associated with the four variants, or three drugs linked to CYP2D6 genotype – tramadol, codeine, or tamoxifen – a clinical decision support system will bring up a pop-up window in the EMR that will alert him to the patient's genotype and make suggestions how to proceed with the prescription.
The biggest barrier to including more sequencing results in the EMR has been the underlying infrastructure. "EMR systems are not designed to handle genomic data, none of them are," Bielinski said. "Essentially, you have to create a way for your genomic data to talk to the EMR – that has not been resolved. This is a big issue across the EMR field, not just at Mayo."
Even putting a modest number of variants into the EMR is a challenge, she said. Unlike blood tests, for example, which change over time so older tests get pushed back, genomic variations are static, and there is no place to put them and make sure they remain easily accessible after several years, she said.
Another problem is scalability. "This initial project is going with four variants, and we can handle four variants, but what happens when you have 100?" Bielinski said. Some EMRs, for example, put genetic variants into the allergy section, which works for small but not for large numbers. "We're still working toward a system where it's more of a seamless scalability," she said.
Finally, it is unclear where and for how long the underlying sequence data should be stored, so it can be reinterpreted in the future when more variants become eligible for inclusion in the EMR. "What obligation do we have, and how do we set up the infrastructure to go re-interrogate the sequence data and pull out anything that may not be clinically actionable now but in six months is?" Bielinksi said. "We haven't solved that at all." This would also have implications for pricing such a test, she said, which would need to factor in the reinterpretation costs.
One question the study will not be able to solve, because the number of patients is too small, is whether so-called preemptive genotyping has advantages over genotyping at the time a drug is prescribed. "In order to look at outcomes, you would have to have a larger sample," Bielinski said. "This initial project was to figure out how to do it," in particular what infrastructure is needed.
However, the project will include surveys of both study participants and doctors who have seen the EMR alerts to find out how they like the system and what concerns they might have. "That's another barrier: If physicians think it's not worth it, then they won't use it," Bielinski said.
To study outcomes, the Mayo researchers are planning to increase the project by enrolling another 3,000 to 5,000 patients from the Mayo Clinic Biobank and have applied for a grant from the National Human Genome Research Institute to fund this effort. The larger study would seek to include more young patients – the current study is tilted towards patients in their 50s or 60s.
Mayo's Center for Individualized Medicine is also interested in offering PGRN-seq as a routine clinical test for doctors to order, although Bielinski provided no timeline for that.
Over the next few years, she anticipates targeted pharmacogenomics tests like PGRN-seq to move on to clinical whole-genome sequencing. "I expect it will become in the very near future cheaper just to do the whole genome than try and do this targeted capture," she said. "The issue then becomes: what to do with the rest of the data?"
Building the infrastructure to store whole-genome data will be "the easiest part" but setting up the informatics and interpretation will be a lot more difficult.
In the future, she said, the Mayo Clinic might be offering clinical whole-genome sequencing along with a menu of tests that patients and their doctors can choose to check off, and only those results would go into the EMR. "People will have a choice about what information they want from their genome," she said.