Genomics in the Exam Room: What's Real, What's Possible, and What You Need to Check First

We're writing this from AIC. Amazing catching up with so many familiar faces and meeting new folks. Some of the best conversations happened between sessions, in the hallways and over coffee. The energy this year was bigger than we could have imagined. One topic came up that we thought would be interesting to explore in this week’s issue: genomic data.

How to use it. When to use it. Which tools are actually safe to run it through.

That last question is the one most clinical AI content skips. We're not going to skip it.

Genomics is genuinely becoming actionable in ways it wasn't even three years ago — pharmacogenomics in particular has moved from specialist report to real prescribing decision. But the infrastructure question underneath it matters just as much as the clinical question. If you're going to bring genetic data into your AI workflows, you need to know exactly where that data is going. So we'll cover both.

98% of your patients carry at least one genetic variant that could change how you prescribe. Most of those patients will never be tested. When a drug does not fit their biology, it looks like bad luck instead of missing data.

Genomic data has been clinically actionable for years. What changed is that AI can finally read it fast enough to use inside a real visit.

Matching a variant to a guideline to a prescription used to be slow, manual work that lived in a specialist report you read once and filed. Now that AI can automate that matching, 76% of health organizations surveyed now run a formal precision medicine program, up from a small minority five years ago. The work has finally moved from the lab report to the exam room.

Here's what that looks like in practice.

Pharmacogenomics is the most mature use of genomic data, and the numbers are hard to argue with. Adverse drug reactions drive 5 to 15 percent of adult hospitalizations. Prescribing guided by a 12 gene panel cut clinically relevant adverse drug reactions by 30%. A patient's CYP2C19 status changes how they metabolize clopidogrel. Their CYP2D6 status affects SSRIs and codeine. SLCO1B1 variants predict statin-induced muscle injury risk. CPIC now publishes genotype based guidance for 79 drugs, covering common prescriptions like clopidogrel, SSRIs, statins, and opioids.

AI is what makes that library usable in a 30 minute slot. A 2025 evaluation of Sherpa Rx, a tool that pairs a language model with CPIC and PharmGKB data, found it outperformed general purpose models like ChatGPT on retrieving and interpreting pharmacogenomic guidance. The clinician asks a plain question about a real patient, and the system returns cited, guideline backed guidance instead of a wall of raw genotypes.

Population guidelines tell you when the average patient should start screening. Polygenic risk scores give you a more personalized timeline. Polygenic risk scores add up the small effects of thousands of variants into a single number that estimates inherited risk for conditions like heart disease, type 2 diabetes, and several cancers. The clinical question was always whether that number adds anything to the risk factors you already measure. The answer is now closer to yes.

In 2025, the European Society of Cardiology endorsed cautious use of polygenic risk alongside traditional tools. Research presented at AHA 2025 showed that adding a polygenic score to the PREVENT cardiovascular tool improved risk prediction across ancestries. AI sharpens these scores further by combining genetics with biomarkers, imaging, and clinical history rather than reading the genome in isolation.

For a longevity or prevention practice, this is the difference between starting screening at a population default age and starting it when a patient's actual inherited risk says to. The score does not replace your judgment. It tells you where to point it.

Hormone therapy is one place where a genetic result can change not just the dose but the route. Hormone therapy raises venous thromboembolism risk roughly threefold, and clotting pathway variants like Factor V Leiden and the prothrombin mutation amplify that risk in the presence of estrogen. Variants in the CYP enzymes that metabolize estrogen can also shift exposure and tolerability.

Read together, that profile helps you decide who is a clear candidate for transdermal rather than oral estrogen, and who needs closer monitoring or a different plan entirely.

AI is useful here because it pulls the clotting and metabolism variants out of a panel and lines them up against the specific formulation you are considering. Be honest with patients that hormone pharmacogenomics is still maturing, so this informs the conversation rather than dictating it.

Clinicians are beginning to use genomic data to frame peptide protocols, but the research is thin and the regulatory picture is still shifting. 

In practice today, clinicians use longevity-associated-variants such as Klotho, FOXO3, and ApoE to frame protocols across diet, lifestyle, and peptide candidates. A possible FDA shift on peptide compounding could also widen the options available to prescribers. Use genomic data in this context as a hypothesis-sharpener. It does not yet settle clinical decisions.

Here's what most genomics content doesn't address, and what you need to think through before your next patient encounter.

Genetic data is not like other clinical data.

A patient's A1c can be re-measured. Their genotype cannot change. It also doesn't only describe them — it describes their biological relatives, including people who never consented to testing. A single DNA sequence reveals information about parents, children, and siblings who never agreed to be in the picture. That's a meaningful distinction when you're deciding what data goes into which system.

"Clinical AI tool" is not a category with a defined standard.

It matters enormously which tool you're using. Even when access controls are in place, LLMs can reproduce identifiable information in generated responses — a model summarizing a patient's data may surface individually identifying details even when that wasn't the intent. Before you run identifiable genomic data through any tool, you need clear answers to three questions: Does the vendor have a signed BAA? What is their data retention and training policy? Is the tool running on a private deployment or sending data to a shared external endpoint?

General-purpose LLMs — in their consumer forms — are not the right tools for identifiable patient genomic data. Purpose-built clinical AI tools with HIPAA-compliant architecture and signed BAAs are a different category. Know which one you're using before you start.

Where to start

Audit your current AI tools for HIPAA compliance before this week's patients. Before you run any genetic data through any tool, confirm you have a signed BAA from the vendor. If you don't know whether you have one, assume you don't and find out. This is not a nice-to-have — it's the floor.

Start with panel data from a clinical lab, not consumer files. A pharmacogenomic panel ordered through a clinical lab has a clear chain of custody, is tied to a patient consent process, and was collected for clinical use. Consumer DNA files have more complicated provenance right now. Start with the cleaner source.

Pick one patient who already has genetic data available to work with. Run it through your clinical AI tool against the medications and risks already in their chart, and look for one thing worth acting on.

Be a modern clinician with the help of Ultralight, the AI-native EHR built specifically for functional, integrative, and longevity medicine.

The Enhanced Games just happened. The Enhanced Games in Las Vegas on May 24 was a one-night competition where athletes openly used testosterone, human growth hormone, and peptides under medical supervision. Only one performance beat a world record, but the cultural moment landed anyway. The company behind it is a DTC telehealth platform already selling personalized testosterone, peptides, and GLP-1s directly to consumers. More patients are going to come in asking about these compounds.

Hims & Hers is coming for the peptide market. Hims & Hers is shifting its business toward wellness and longevity, with peptides as the next big bet. The company already owns a peptide manufacturing facility in California and is waiting on FDA regulatory clarity to scale. They built a massive patient base selling GLP-1s over the internet. Peptides are next. 

Oura Ring 5 is here, and it's 40% smaller. Oura just announced the Ring 5, shipping June 4 at $399. It's 40% smaller than Ring 4, adds blood pressure monitoring, and introduces GLP-1 insights and lab upload features. More patients are going to show up to appointments with continuous biometric data already in hand. 

Whole genome sequencing is moving into preventive care. Illumina and Veritas launched a preventive genomics consortium in March 2026 aimed at bringing whole genome sequencing into routine care through health insurance plans. If sequencing becomes a covered baseline, the question shifts from whether your patients have genomic data to whether your practice can act on it.

Upcoming Conferences & Events

Jun 9–11, Longevity Docs Cannes 2026 · Cannes, France · Invite-leaning room for clinicians at the frontier of longevity medicine. Worth it if you are designing the next iteration of your own practice.

Sep 24–26, Vibrant Longevity Summit · Austin, TX  · A clinical room of practitioners running lab-driven, longitudinal care. For anyone building a practice around diagnostics and biomarkers who wants peers who work the same way.

Oct 8–10, A4M Women's Health Summit · San Antonio, TX ·  The best clinical education on hormone, metabolic, and midlife women's health you will see this year. The room to be in if you are growing the perimenopause and menopause side of your practice.

Oct 21–24, NAMS Annual Meeting · San Diego, CA  · The single most practice-changing meeting of the year for midlife women's health. Your protocols will look different after this one.

Nov 5–8, Eudēmonia Summit · West Palm Beach, FL ·  One of the most talked-about longevity gatherings in the U.S. Experientials, hands-on demos, and the best place to try the emerging frameworks your patients will ask you about next year. Ovation and Ultralight team will be there!

Nov 5-7 — Private Physicians Alliance Annual Meeting · St. Petersburg, FL The gathering for independent, cash-pay, and concierge physicians navigating practice independence. Practical and peer-driven. Ultralight will be there!

Nov 8-11 — American College of Lifestyle Medicine Conference · Orlando, FL Lifestyle medicine's main annual event — evidence-based approaches to behavior change, chronic disease, and healthspan. Growing overlap with the longevity medicine community.

Dec 11–13, A4M Longevity Fest · Las Vegas, NV  · The biggest longevity event in the U.S. The room spans clinicians, industry, founders, and the people building next year's platforms, and the connections from this one tend to compound through the rest of your year. Ultralight will be there!

Know of an event we should add? Reply and tell us.

If you're already running genetic data through clinical AI tools and you've worked out a protocol that holds up — clinically and on the data handling side — reply and tell us what you're doing. That's the kind of thing worth sharing.

Did you attend AIC this year? Any topics you want us to cover? The best ideas in this newsletter come from clinicians doing the work.

Until next week, keep building the practice you imagined when you started. We are building it with you.

— Sunita and Dr. G