Pharmacogenomic Testing at Home: What You Can Call Yourself From Your Own Genome

You can get pharmacogenomic results at home two ways. Buy a saliva kit from a PGx vendor, which genotypes a fixed panel of star alleles and mails back a PDF grouping drugs into green/yellow/red bins. Or sequence your genome once and call the pharmacogenes yourself with open-source tools, which gives you the underlying diplotypes, the evidence level behind each call, and the ability to re-run the analysis when CPIC updates a guideline. We prefer the second, with one caveat: whatever you produce at home is a research-grade result. Before any prescribing decision is made on it, the relevant variant should be confirmed in a CLIA/CAP laboratory and interpreted by a clinician or pharmacist. Nothing below is medical advice, and no part of it should be used to start, stop, or change a medication.

What the panels genotype, and what they miss

A consumer PGx kit is usually a targeted genotyping assay: a fixed set of probes or a mass-spectrometry panel that interrogates a few dozen to a few hundred positions across the following genes.

• CYP2C19
• CYP2D6
• CYP2C9
• CYP3A5
• SLCO1B1
• VKORC1
• TPMT
• NUDT15
• DPYD
• UGT1A1
• a handful of others

Targeted mass-spec assays are fast and cheap per sample and reproduce well in clinical workflows ¹. The limitation is structural: you only see the positions the assay was designed for. A variant not on the panel is silently called wild type, which is how you get a confident “normal metabolizer” that is wrong.

Sequencing changes the denominator. Caspar and colleagues profiled pharmacogenes from whole-genome data and found variants of potential pharmacogenetic relevance that standard targeted panels do not interrogate, including rare and structural variation ². The core argument for doing PGx off a genome rather than off a chip rests on this. The same 30x BAM that you already have covers every pharmacogene, and you can re-interrogate it in five years when a new star allele is defined.

Three concrete misses that matter:

• DPYD HapB3 (c.1129-5923C>G) is deep intronic. Several panels include the common *2A splice variant and c.2846A>T but not HapB3.
• CYP2D6 copy number. Deletions (*5), duplications, and CYP2D6–CYP2D7 hybrids are not detectable by SNP genotyping alone. A duplicated *1 and a single *1 are different phenotypes.
• Rare loss-of-function alleles in TPMT and NUDT15 that are population-specific and simply absent from panels designed on European allele frequencies.

The pipeline we would run

Start from short-read WGS at 30x or better, aligned to GRCh38, with a joint-called or single-sample VCF. Exome data works for some genes and fails for DPYD HapB3 and for anything intronic or regulatory, including VKORC1 -1639G>A if the capture does not extend upstream.

Step 1: PharmCAT. PharmCAT takes a VCF, matches named alleles, assigns diplotypes, maps them to phenotypes, and writes a report keyed to CPIC and DPWG guidance. It is strict about input, which is a feature. The preprocessor normalizes your VCF against the exact positions PharmCAT needs:

pharmcat_pipeline sample.GRCh38.vcf.gz \
  -o pharmcat_out \
  --missing-to-ref

The --missing-to-ref flag tells PharmCAT to assume any position absent from your VCF is homozygous reference, so never use it blindly. That assumption is safe only if you are working from a gVCF or from a VCF you know includes callable-region information, so that “absent” means “called reference” rather than “never sequenced”. If you pass a sites-only variant VCF from a pipeline that drops reference calls, you cannot distinguish the two, and every uncovered position becomes a false wild-type call. Run once without the flag and read the “missing positions” section of the report first.

Step 2: CYP2D6 separately. PharmCAT does not call CYP2D6 from a VCF, because the locus needs read-depth and structural analysis. Use Aldy or StellarPGx on the BAM:

aldy genotype -p wgs -g CYP2D6 sample.bam

Aldy models copy number and hybrid configurations explicitly. Check the reported confidence and the coverage profile across CYP2D6 and CYP2D7, and expect ambiguity: the gene is GC-rich, the paralog is 94% identical, and short reads do not always resolve which copy carries which variant. If CYP2D6 matters to you specifically, long-read sequencing or a validated clinical assay is the right answer. A harder squint at 150 bp reads is not. Feed the resulting diplotype back into PharmCAT as an outside call so the report includes it.

Step 3: HLA. HLA-B*57:01 and HLA-A*31:01 carry some of the strongest pharmacogenomic associations in the CPIC catalog. Standard variant callers do not type HLA. Run a dedicated typer such as HLA-LA over the BAM and treat the result as a hypothesis. The result is not a call. HLA typing from short reads is good at two-field resolution for common alleles and less good at rare ones. A clinical HLA-B*57:01 test exists and is cheap. If this allele is relevant to you, get the clinical test.

Step 4: Read the report past the colors. PharmCAT’s output distinguishes CPIC evidence levels. A Level A recommendation with a strong mechanism (DPYD and fluoropyrimidines, TPMT/NUDT15 and thiopurines, CYP2C19 and clopidogrel) is a different kind of statement than a gene-drug pair where the evidence is thin and the panel vendor has bundled it in anyway.

What the evidence supports, and what it does not

Most consumer PGx marketing is built on psychiatry, where the claims outrun the data. A real-world analysis of patients with major depressive disorder found that PGx-informed care changed prescribing patterns and healthcare resource use ³, and a quality-improvement project in depression care reported similar directional findings ⁴. These are meaningful but modest effects, and the psychiatric PGx field has a well-documented gap between what commercial panels assert and what the underlying evidence establishes ⁵. Treat any report that ranks antidepressants into tiers as a prompt for a conversation with a prescriber. Do not treat it as a ranking.

The stronger case is in oncology, transplant, and pain, where a single variant can change a drug’s exposure by an order of magnitude. Implementation studies in pediatric tertiary care found actionable pharmacogenomic results in a large fraction of sequenced patients ⁶, and community-hospital deployments have shown that pharmacist-led review of PGx results changes real prescriptions ⁷.

The recurring problem in every implementation paper is getting the result to the prescriber at the moment of prescribing, in a form the EHR can act on ^{8 9}. The assay is not the problem. Interpretation also drifts: the same sequence data can yield different diplotype calls depending on allele definition version and caller, which is why clinical laboratories version their calls ¹⁰. If you generate results at home, you own that problem. Export the raw diplotypes, record the tool versions and the CPIC allele definition release, and hand the clinician the diplotype and the evidence level rather than the vendor’s color code.

Questions people also ask

Does insurance pay for pharmacogenetic testing? Sometimes, and coverage is inconsistent by payer, indication, and gene. Single-gene tests with strong guideline support ordered before a specific drug (DPYD before fluoropyrimidines, TPMT/NUDT15 before thiopurines, HLA-B*57:01) are covered more often than multi-gene psychiatric panels. Reimbursement uncertainty is one of the named barriers to clinical PGx adoption ^{8 5}. Ask the ordering clinician’s office to run a benefits check with the specific CPT codes before the sample is collected.

How do I get GeneSight testing covered by insurance? It requires a clinician’s order and, for most payers, documentation of prior medication trials that did not work or were not tolerated. Coverage policies for combinatorial psychiatric panels are narrower than for single-gene testing. The prior-authorization request comes from the prescriber. It does not come from you.

How do I get pharmacogenetic testing? Three routes: a clinician orders it through a clinical laboratory, a direct-to-consumer vendor ships you a saliva kit with physician review attached, or you sequence your genome and call the pharmacogenes yourself. The first is the only one whose result goes into your chart as a clinical finding without re-testing.

Can I use my existing 23andMe or ancestry raw data for PGx? Partially, and carefully. Array data covers a subset of common star alleles, cannot resolve CYP2D6 copy number, and has no reference-call information, so missing positions are genuinely unknown. PharmCAT will accept a converted VCF, but you should read its missing-position report before you read anything else.

What does a home result change today? On its own, nothing about your prescriptions. It gives you a durable, versioned record of your diplotypes that a pharmacist or prescriber can confirm and act on. Any change to a medication is their decision, made with the clinical context you have and a text file does not.

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. Sylvan M. Caspar, Timo Schneider, Janine Meienberg, et al. Added Value of Clinical Sequencing: WGS-Based Profiling of Pharmacogenes. International Journal of Molecular Sciences, 2020. https://doi.org/10.3390/ijms21072308 ↩

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. Jeremy Tyler Rockwell, Amy Y. Spurlock, Kerri L. Outlaw. Pharmacogenetic Testing in Patients With Depression: A Quality Improvement Project. Journal of the American Psychiatric Nurses Association, 2026. https://doi.org/10.1177/10783903261431278 ↩

5. 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

6. Iris Cohn, Roozbeh Manshaei, Eriskay Liston, et al. Assessment of the Implementation of Pharmacogenomic Testing in a Pediatric Tertiary Care Setting. JAMA Network Open, 2021. https://doi.org/10.1001/jamanetworkopen.2021.10446 ↩

7. Jingyi Wang, Sue Paul, Renee Noel Arbet, et al. Application of Pharmacogenomics Testing in a Community-based Facility. Hospital Pharmacy, 2022. https://doi.org/10.1177/00185787221134693 ↩

8. Ann M Moyer, Pedro J Caraballo. The challenges of implementing pharmacogenomic testing in the clinic. Expert Review of Pharmacoeconomics & Outcomes Research, 2017. https://doi.org/10.1080/14737167.2017.1385395 ↩ ↩²

9. Paul C.D. Bank, Jesse J. Swen, Henk-Jan Guchelaar. Implementation of Pharmacogenomics in Everyday Clinical Settings. Advances in Pharmacology, 2018. https://doi.org/10.1016/bs.apha.2018.04.003 ↩

10. Yuan Ji, Yue Si, Gwendolyn A. McMillin, et al. Clinical pharmacogenomics testing in the era of next generation sequencing: challenges and opportunities for precision medicine. Expert Review of Molecular Diagnostics, 2018. https://doi.org/10.1080/14737159.2018.1461561 ↩