NEW YORK (GenomeWeb) – A consortium of researchers has developed a bioinformatic software tool for the annotation and interpretation of pharmacogenetic variants that it hopes will improve the implementation of pharmacogenomics in the clinic.
The software, called Pharmacogenomics Clinical Annotation Tool (PharmCAT), takes PGx variants from sequencing and genotyping data of certain genes, interprets the variant alleles, infers haplotype pairs and star alleles, and generates guideline-based reports that can help physicians make drug prescription decisions.
The team, led by Teri Klein at Stanford University and Marylyn Ritchie at Geisinger Health System, plans to present PharmCAT next month at the American Society of Human Genetics annual meeting in Vancouver and to make a first version available to researchers as an open-source tool by the end of October.
Pharmacogenetic haplotypes are usually described by the so-called star allele nomenclature, which uses the gene name, a star, and a number and/or letter rather than the genomic position of a variant. One difficulty in using DNA sequence or genotyping data for pharmacogenomic testing has been the translation of genomic variants into this nomenclature.
According to Ritchie, who chairs the department of biomedical and translational informatics at Geisinger, the main idea of developing PharmCAT has been to standardize the process of making star allele calls. Automated PGx variant annotation is currently pursued by many clinical and research laboratories, she said, which "are all writing little scripts and pieces of software that take either genotype files or variant files and translate them into these star alleles or haplotypes," which then allow the application of guidelines around gene-drug pairs that were developed by the Clinical Pharmacogenetics Implementation Consortium (CPIC).
Because of a lack of standardization, she said, the process sometimes leads to different results between labs. For example, some consider rare alleles in the interpretation that not all assays measure, while others don't. "You should not have one clinical lab run an [Affymetrix] pharmacogenomics assay and then another one run an Illumina assay and give you different results, one saying that you are a poor metabolizer and one saying you are a standard metabolizer," Ritchie said. "That should not happen, and that is happening right now."
This, in turn, could lead to bad prescribing decisions. "We worry that if people aren't calling these alleles right and thus not using the CPIC guidelines appropriately, that they will be making the wrong recommendations," Ritchie said. "Then it loses the accuracy that we all think pharmacogenomics will have."
PharmCAT currently only looks at variants from CPIC level A genes, with the exception of G6PD, HLA, and a few complex CYP2D6 alleles that are difficult to distinguish from pseudogenes. Those three genes require long-read sequencing data to interpret them well, Ritchie said, which most institutions do not have available at the moment.
For CPIC level A genes and their drugs, genetic information should be used to prescribe the drug, and a high level of evidence exists for doing so. PharmCAT generates a report that is based on clinical recommendations from the CPIC guidelines. This report will not only say which variant alleles were interpreted but also which ones could not be assessed because they were not part of the input data, Ritchie said.
The tool has been developed in a collaboration between the Pharmacogenomics Research Network (PGRN) Statistical Analysis Resource (P-STAR), the Pharmacogenomics Knowledgebase (PharmGKB), the Clinical Genome Resource (ClinGen), CPIC, and the Electronic Medical Records and Genomics (eMERGE) Network. The hope is that a tool developed by many members of the PGx community will be adopted by a lot of laboratories, Ritchie said.
In April, a team of programmers got together for a week-long "hackathon" at PharmGKB at Stanford University to develop a prototype, Ritchie said. PharmGKB researchers have refined and cleaned this up, leading to the current version, which is posted in GitHub and is technically still an alpha version.
Programmers at Geisinger and other institutions are now testing the tool prior to the formal release of version 1, along with a manuscript, at the end of next month.
One dataset they are using to test PharmCAT is CLIA-validated genotypes from the eMERGE PGx project, where particular star alleles have already been confirmed. "Those are a great test set to make sure the software is working," Ritchie said. "We should pick up the same things that they picked up manually."
However, the eMERGE researchers often only looked at one or two genes, and PharmCAT would now allow them to study all 84 genes that were analyzed as part of that project.
The first version of PharmCAT will not yet be ready for clinical use, which might require additional quality control elements, Ritchie said, but once the tool is out, the developers plan to interact with clinical labs to see what is required. Overall, she said, PharmCAT could be useful to any research or clinical laboratory that needs to derive pharmacogenomic haplotypes from genomic variant data.
The Genetic Testing Laboratory at the Icahn School of Medicine at Mount Sinai is one of the labs interested in testing PharmCAT. "It's something we would consider and test for inclusion in our informatics workflow," along with in-house developed tools, said Stuart Scott, director of Mount Sinai's Molecular Genetics Laboratory as well as its Cytogenetics and Cytogenomics Laboratory. Scott has also been involved in the development of PharmCAT.
Mount Sinai's lab has been offering pharmacogenetic testing for seven or eight years now, he said, using a clinical targeted genotyping test from Luminex that comes with built-in software to infer the patient's star allele diplotype. About 2,000 patients have received this test as part of an ongoing study, and the researchers are now looking into how physicians use the information to prescribe medications.
But the lab has been developing a broader next-gen sequencing-based PGx panel that includes more than 30 genes, which it is currently validating, and PharmCAT could be useful for that. The new panel consists of a set of clinically more conservative, actionable genes and an expanded set of genes where the evidence may not be high enough to change clinical decision-making but would still be of interest to informed physicians.
"Once we have our full-gene next-gen sequencing PGx panel deployed, that's when we need a tool like PharmCAT, because you need a much higher-level sophisticated informatics tool that can go from the full gene VCF file back to the star allele diplotype system," Scott said.
In addition, he said, the tool might be helpful generally for clinical research projects that involve exome sequencing or other types of high-throughput sequencing. "Pharmacogenetic results are often some of the first that are considered to be returned to patients or research subjects in clinical genomic medicine studies," he said.
Proper validation of the PharmCAT, which is ongoing, will be particularly important, he said. Mount Sinai is also involved in this, testing DNA samples that have been genotyped by its existing clinical platform and making sure PharmCAT delivers the same results.