Pharmacogenomics Companies, and What You Can Do With Your Own Genome Instead

The companies most people encounter are GeneSight (owned by Myriad Genetics, which acquired Assurex Health in 2016), OneOme (co-founded by Mayo Clinic), Genomind, Tempus, and the large reference labs Quest Diagnostics and Labcorp, alongside dozens of smaller CLIA labs such as Admera and Coriell. Nearly all of them run the same underlying assay: a targeted genotyping panel covering a few dozen predefined variants in fifteen to thirty pharmacogenes, interpreted against published guidelines. If you already have whole-genome sequencing, you can generate the same star-allele calls yourself with open-source tools, and cover variants the panels do not carry. Deciding what to take, change, or stop is a prescribing decision and belongs with the clinician who knows your history. You cannot make that decision yourself.

What these companies run

A pharmacogenomic (PGx) test is almost always a targeted genotyping assay. It is almost never sequencing. The lab picks a fixed set of positions, types them on a microarray, a MassARRAY-style mass spectrometry platform, or a small amplicon panel, and translates the genotype into star alleles (CYP2C19*2, CYP2D6*4) and then into a metabolizer phenotype. Mass spectrometry genotyping in particular has become a workhorse for clinical PGx because it multiplexes dozens of SNPs cheaply with high call rates ¹.

The differences between vendors are mostly in three places. First, the gene list. A minimal panel covers CYP2C19, CYP2C9, CYP2D6, CYP3A5 and TPMT. It also covers NUDT15, DPYD, SLCO1B1, VKORC1 and UGT1A1. Broader panels add CYP2B6, CYP1A2, CYP4F2 and COMT, along with HTR2A and HLA alleles. Second, the depth within each gene: two labs can both claim “CYP2D6” and one types four alleles while the other types thirty plus copy number. Third, and most consequential, the interpretation layer. Most labs report against CPIC (Clinical Pharmacogenetics Implementation Consortium) guidelines and the FDA’s table of pharmacogenetic associations, which is a transparent, citable mapping. GeneSight and Genomind instead report a proprietary combinatorial algorithm that weighs several genes together into red/yellow/green medication bins, and the weighting is not public.

We prefer the CPIC-style reporting, for a simple reason: you can audit it. Given a star-allele diplotype, you can look up the CPIC guideline, read the evidence level, and see exactly why a gene-drug pair is annotated the way it is. A proprietary bin gives you a conclusion without the derivation.

What the evidence supports

Single-gene pharmacogenetics with strong mechanistic backing is on firm ground. DPYD variants and fluoropyrimidine toxicity, TPMT and NUDT15 with thiopurines, HLA-B*57:01 with abacavir, and CYP2C19 with clopidogrel activation are well-replicated associations with dosing guidance written by professional bodies.

Combinatorial psychiatric panels are a different matter. A meta-analysis of randomized trials in major depressive disorder found that pharmacogenomic-guided care was associated with improved response relative to treatment as usual, but the trials were largely open-label, industry-sponsored, and used remission and response scales that are sensitive to unblinding ². More recent real-world claims analyses report changes in medication use and healthcare resource utilization following testing ³. That tells you that prescribing behavior changes. It does not tell you that outcomes improve for the reason the test proposes. A critical review of adoption barriers in psychiatry makes the same point about the gap between analytic validity (the test calls the genotype correctly) and clinical utility (acting on it helps) ⁴. Our reading: treat combinatorial psychiatric reports as a weak prior, not evidence, and treat the individual CPIC-level gene-drug pairs inside them as the part worth attention.

The industry-side view is similar. The Industry Pharmacogenomics Working Group has catalogued the recurring obstacles to embedding PGx in drug development and practice, including inconsistent variant nomenclature, incomplete allele coverage across platforms, and reimbursement uncertainty ⁵. Those problems are twenty years old. An early assessment of genome-based pharmacogenetics in pharma anticipated exactly this pattern of slow translation from association to prescribing practice ⁶.

Doing it yourself from whole-genome data

If you have a 30x whole genome as a CRAM or BAM plus a joint-called VCF, you can produce most of a commercial PGx report in an afternoon. Whole-genome sequencing also finds rare and novel variants in pharmacogenes that no fixed panel carries, which matters because a meaningful share of functionally relevant pharmacogene variation is rare rather than common ⁷.

Start with PharmCAT, which takes a VCF and emits star-allele diplotypes, CPIC phenotypes, and a report:

pharmcat_pipeline sample.GRCh38.vcf.gz -o pharmcat_out --reporter-html

Three failure modes matter here. PharmCAT requires GRCh38 coordinates, so a GRCh37 VCF must be lifted over (bcftools +liftover or CrossMap) before anything else. It also needs to know reference calls. Variant sites alone are not enough. If your VCF contains only ALT positions, PharmCAT cannot distinguish “reference” from “not sequenced” and will return uncertain calls. The preprocessor’s --missing-to-ref flag forces missing positions to reference, which is only safe if you have verified coverage there. Check first:

samtools depth -a -r chr10:94759000-94858000 sample.cram | awk '$3<10' | wc -l

That interval is CYP2C19. Do the same for CYP2C9, DPYD, and TPMT. Any gene with a coverage hole gets excluded from your interpretation rather than silently called reference.

CYP2D6 is the hard case and PharmCAT will not call it from a VCF. The gene sits next to the CYP2D7 pseudogene, undergoes whole-gene deletion and duplication, and forms CYP2D6-CYP2D7 hybrid alleles. Use a caller built for it. Cyrius reads the alignment directly:

star_caller.py --manifest cram_manifest.txt --genome 38 --prefix sample --outDir cyp2d6_out --threads 8

Aldy is a good cross-check and covers more genes, including copy-number-aware calls on CYP2D6, CYP2A6, and others:

aldy genotype -p wgs -g CYP2D6 --genome hg38 sample.cram

When Cyrius and Aldy disagree, the usual cause is a structural rearrangement that one model resolves and the other does not, and the correct response is to report the genotype as ambiguous. Feed the resolved diplotype back into PharmCAT with --matcher-... outside calls (PharmCAT accepts an outside-call TSV) so the final report is internally consistent.

For HLA alleles relevant to hypersensitivity (HLA-B*57:01, HLA-B*15:02, HLA-A*31:01), WGS-based typing with HLA*LA or arcasHLA is accurate enough for screening but is not clinical-grade typing. If an HLA call would inform a real prescribing decision, it needs confirmation in a clinical lab.

One variant deserves specific mention because panels vary: the DPYD HapB3 allele is tagged by a deep intronic variant, rs75017182 (c.1129-5923C>G), which many array-based panels omit entirely. Whole-genome data captures it. Whether it changes anything about a fluoropyrimidine regimen is a decision for an oncology team. It is not a decision for you and a script.

Questions people also ask

What company owns GeneSight? Myriad Genetics. Myriad acquired Assurex Health, the developer of GeneSight, in 2016, and sells the test through its own lab.

Is GeneSight FDA approved? No. GeneSight is a laboratory developed test run in a CLIA-certified, CAP-accredited lab, which means it is regulated as a lab service rather than cleared or approved as a device. The FDA issued a safety communication in 2018 warning that some pharmacogenetic tests make drug-specific claims not supported by evidence, and that remains the relevant context for interpreting combinatorial reports.

Is GeneSight trustworthy? Its analytic validity, meaning whether it calls the genotypes it claims to call, is not the issue in dispute. The dispute is clinical utility: the randomized evidence for combinatorial psychiatric panels is mostly open-label and sponsored, and independent reviews have been cautious ²⁴.

Which insurance covers pharmacogenomic testing? Medicare covers some psychiatric PGx testing under MolDX local coverage determinations for specific indications, while many commercial plans still classify combinatorial panels as investigational. Coverage for sequencing-based testing generally has lagged the technology by years ⁸. Verify with your specific plan before assuming anything.

Does GeneSight sell your information? Myriad is a HIPAA-covered clinical lab, which is a stricter regime than consumer genetics companies operate under, and its policies permit use of de-identified data for research. Read the current privacy notice and authorization forms you sign rather than relying on a summary.

Which companies provide pharmacogenomic testing? GeneSight (Myriad), OneOme, Genomind and Tempus are examples. Others are Quest Diagnostics, Labcorp, Admera, and Coriell, among many regional CLIA labs. They differ more in gene coverage and interpretation policy than in underlying assay technology, so compare the published allele lists before comparing prices.

Footnotes

1. Liqian Mo, Xin Luo, Caihua Yang, et al. Current status and prospects of nucleic acid mass spectrometry in clinical pharmacogenomics. Precision Medication, 2024. https://doi.org/10.1016/j.prmedi.2024.10.001 ↩

2. Joshua D. Rosenblat, Yena Lee, Roger S. McIntyre. Does Pharmacogenomic Testing Improve Clinical Outcomes for Major Depressive Disorder?. The Journal of Clinical Psychiatry, 2017. https://doi.org/10.4088/jcp.15r10583 ↩ ↩²

3. Andria L. Del Tredici, Holly L. Johnson, Brady DeHart, et al. Real-World Impact of Pharmacogenomic Testing on Medication Use and Healthcare Resource Utilization in Patients With Major Depressive Disorder. Journal of Clinical Psychopharmacology, 2025. https://doi.org/10.1097/jcp.0000000000001999 ↩

4. Catherine R. Virelli, Ayeshah G. Mohiuddin, James L. Kennedy. Barriers to clinical adoption of pharmacogenomic testing in psychiatry: a critical analysis. Translational Psychiatry, 2021. https://doi.org/10.1038/s41398-021-01600-7 ↩ ↩²

5. Karina Bienfait, Aparna Chhibber, Jean-Claude Marshall, et al. Current challenges and opportunities for pharmacogenomics: perspective of the Industry Pharmacogenomics Working Group (I-PWG). Human Genetics, 2021. https://doi.org/10.1007/s00439-021-02282-3 ↩

6. Allen D. Roses. Genome-based pharmacogenetics and the pharmaceutical industry. Nature Reviews Drug Discovery, 2002. https://doi.org/10.1038/nrd840 ↩

7. Clint Mizzi, Brock Peters, Christina Mitropoulou, et al. Personalized Pharmacogenomics Profiling Using Whole-Genome Sequencing. Pharmacogenomics, 2014. https://doi.org/10.2217/pgs.14.102 ↩

8. Shraddha Chakradhar. Insurance companies are slow to cover next-generation sequencing. Nature Medicine, 2015. https://doi.org/10.1038/nm0315-204 ↩