Herasight vs Orchid: what actually separates two embryo screening companies

Five euploid embryos. Two screening companies. Both will rank your embryos by genetic disease risk. Both will hand you a report with numbers that look precise. The numbers aren't measuring the same thing.

Herasight and Orchid are both CAP-accredited and CLIA-certified, with labs in North Carolina. Both offer preimplantation genetic testing for polygenic conditions (PGT-P) with counseling included. MIT Technology Review named embryo scoring one of its 10 Breakthrough Technologies for 2026. On a spec sheet, these two companies look interchangeable.

They aren't. Orchid has published within-family validation data for 6 of the 12 diseases they screen as standard. Herasight has published within-family validation for all 17 disease scores in Moore et al. 2025. On conditions where both companies have data, Herasight's scores explain roughly twice as much of the variation in disease risk between sibling embryos.

The gap isn't a branding difference. Orchid is a sequencing company; polygenic scoring is a product layered on top of their sequencing pipeline, and there's no statistical geneticist on the team. Herasight is built around the statistical genetics of polygenic prediction, with scores designed from the ground up for embryo screening. That difference shows up in almost every question worth asking.

Why sibling validation changes the comparison

Here's why this is the question most families never think to ask.

When a polygenic score separates one stranger from another, it draws on every source of genetic difference between them. Some of those differences are actual variants that drive disease risk. Others are ancestry patterns and shared environments that have nothing to do with the disease itself. For clinical medicine, where a doctor estimates one patient's risk relative to the general population, that's fine.

Embryo screening is a completely different situation. You're not comparing strangers. You're choosing between embryos that all came from the same two parents. They share roughly half their genetic variation. The ancestry patterns and family environment are held constant. A score that leans on any of those shared sources of variation will lose some apparent accuracy when applied to siblings.

How much accuracy gets lost depends on the trait. For something like educational attainment, where family environment and population structure play a large role, the drop is substantial. For most disease traits, the attenuation tends to be small. But "probably small" isn't the same as "tested and confirmed for every condition you're being screened for." That's the gap.

The question that matters for your IVF decision isn't "can this score tell a random person apart from another random person?" It's "can this score tell these specific embryos apart from each other?" The only way to answer that is to test the score on siblings.

What the data shows

Moore et al. 2025 tested this directly. Herasight evaluated 17 disease polygenic scores on sibling pairs in large biobank datasets. Sixteen of 17 showed no decrease in predictive performance within families compared to population-level predictions. The one exception was osteoporosis, which showed a modest attenuation. It's still predictive between siblings, just less sharply so. Overall, the result means these scores aren't riding on population structure artifacts. They're picking up genuine causal genetic signal that actually differs between one embryo and the next.

Orchid has also published within-family analysis in a technical guide, using UK Biobank sibling pairs with imputed parental genotypes. Their results show direct-to-population effect ratios of 0.92 to 1.04 across seven conditions. That's reassuring for those specific diseases. But of the 12 conditions Orchid screens for as standard, only 6 are covered by that analysis. Schizophrenia and bipolar disorder were excluded for insufficient case counts. Type 1 diabetes, celiac disease, and Alzheimer's disease were excluded as oligogenic. Those are all conditions Orchid screens for but hasn't validated within families.

The more consequential gap is absolute predictive accuracy. On conditions where both companies have within-family data, Herasight's scores explain roughly twice as much of the variation in disease risk between sibling embryos. That variation is what an embryo score is trying to capture. Catching more of it gives you a more reliable ranking of the embryos you actually have.

Think about what that means for a real family. Say both parents carry significant genetic risk for type 2 diabetes. They have five chromosomally normal embryos after PGT-A. In families where both parents are affected, the absolute risk difference between the highest-risk and lowest-risk embryo can exceed 40%. How reliably you can identify which embryo is which depends on the score. Scores that explain more of the variation give you a more confident ranking.

How each lab actually builds the test

Orchid does deep whole-genome sequencing on every embryo. One biopsy, one workflow, one report covering aneuploidy, single-gene conditions, and polygenic scoring. The sequencing itself is strong for what it does, which is mostly monogenic detection and structural coverage.

Herasight's workflow is different on purpose. Parents get 30x whole-genome sequencing using both short and long reads. Embryos get 2x low-pass sequencing. The parental data is then used to reconstruct the embryo genomes at higher effective resolution than per-embryo sequencing typically provides. This trio approach is more accurate for polygenic prediction because embryo DNA samples are small and noisy, while parental samples are clean and plentiful.

Herasight also runs aneuploidy, monogenic, and polygenic screening in a single workflow. That isn't an Orchid-only capability. And Herasight's aneuploidy screening runs at roughly ten times the coverage depth Orchid uses.

The one thing per-embryo deep sequencing does better is detect de novo mutations that appear in the embryo but not in either parent. That's a narrow clinical need. For families who want it, Herasight offers a higher-coverage sequencing add-on (~40x) on a single embryo after the initial screen narrows down candidates. Running it on every embryo is expensive and almost always unnecessary.

There's also a Herasight capability with no Orchid equivalent. If your clinic has already run PGT-A on your embryos, Herasight's ImputePGTA reconstruction can pull polygenic risk scores from that existing PGT-A data plus parental sequencing. No new embryo biopsy required. Orchid can't do that.

The ancestry and trait-breadth gap

For non-European families, there's another dimension that can matter more than methodology.

Orchid publishes its own ancestry table. Families of European descent receive all 12 polygenic reports. South Asian families receive 10. East Asian families receive 9. Families of African descent receive 2. Orchid delivers a report for a given disease only when the score's accuracy clears a strict threshold for that ancestry, and for most diseases and most non-European ancestries, that threshold isn't met.

This is a field-wide problem, not specific to Orchid. Most polygenic scores were built from datasets that skew heavily European, and scores trained primarily on European data perform substantially worse for other populations. It's been one of the biggest limitations of polygenic screening from the beginning.

Herasight calibrates and validates scores across 8+ ancestry groups. Moore et al. 2025 uses methods including SBayesRC, a newer statistical approach that uses functional genomic annotations to improve how well scores transfer between populations. For a non-European family, the comparison often narrows to whether Orchid can deliver anything useful at all. For a family of African descent, the gap between 2 conditions and Herasight's full trait catalog is the whole comparison.

Even for European families, the catalogs don't line up. Orchid's 12 standard diseases are Alzheimer's, atrial fibrillation, bipolar disorder, breast cancer, celiac disease, coronary artery disease, inflammatory bowel disease, prostate cancer, severe obesity, schizophrenia, type 1 diabetes, and type 2 diabetes. Herasight covers all 17 scores validated in Moore et al. 2025 plus additional conditions Orchid doesn't offer as standard, including ADHD and cognitive ability screening. Cognitive screening in particular is a Herasight product with no Orchid equivalent.

If your main concern is a specific single-gene condition, both companies can screen for it alongside aneuploidy and polygenic scoring in the same workflow. If your concern is polygenic accuracy, the validation and R² gaps are what show up in practice: more confident rankings, wider trait coverage, and no ancestry-based gating. If you're a non-European family, check the trait-by-ancestry list before anything else. That's often where the real comparison ends.

And if you've already run PGT-A, Herasight can pull polygenic results from that existing data. Orchid can't. Herasight's carrier screening also covers more monogenic conditions, and parental sequencing is standard rather than an add-on.

Both companies are self-pay with comparable logistics. The science underneath the polygenic predictions is where they come apart.

If you want to see what the numbers look like for your specific family situation, try the calculator. If you have questions about what this comparison means for you, reach out.

Herasight does not provide medical diagnoses or tell patients which embryo to transfer. Polygenic risk scores are probability estimates, not guarantees. The decision of which embryo to transfer is between you and your clinician.