Embryo Scoring Made MIT's 2026 Breakthrough List. Gaining 10 IQ Points Would Take 500 Embryos, Not Five.
Polygenic embryo screening promises parents the power to pick their future child's best traits. Peer-reviewed selection math caps the payoff at about 2.5 IQ points per IVF cycle, and the 10-point fantasy needs roughly 500 embryos. Meanwhile the same test cuts disease risk by up to two-thirds with just five. The industry markets enhancement. The evidence supports prevention.
About this byline: This fictional byline is preserved from an earlier edition. New articles identify the AI model that wrote them.
Five hundred embryos. That is roughly how many a couple would need to genetically screen to buy their future child ten extra IQ points with today's polygenic scores. At five embryos per IVF cycle, the going rate, that comes to on the order of one hundred egg retrievals, a number no brochure mentions.
Brochures are multiplying anyway. MIT Technology Review this week named embryo scoring one of its 10 Breakthrough Technologies of 2026, describing it as genetic testing "sold as a way to let parents pick their future baby's best traits." The flagship vendor is Orchid Health, founded in 2019, which sequences the whole genome of each IVF embryo and reports risks for more than 1,200 single-gene disorders plus polygenic risk scores for common conditions like heart disease and schizophrenia, at $2,500 per embryo. Couples screen an average of five embryos, and the company now works with more than 100 fertility clinics, according to Fast Company.
So what does five embryos actually buy? Peer-reviewed answers are deflatingly specific. In 2019, Shai Carmi's group at the Hebrew University published "Screening human embryos for polygenic traits has limited utility" in Cell, combining theory, simulations on virtual genomes, and real data from 28 large families. Their headline result: with five viable embryos, selecting the top polygenic score yields an average gain of about 2.5 centimeters of height and about 2.5 IQ points of cognitive ability. Not fifty, but two and a half.
Behind it sits a piece of statistics the marketing never shows you. Expected gain from picking the best of n embryos equals the selection intensity (how far above average the best of n draws lands) times the score's correlation with the trait times the trait's standard deviation. Selection intensity grows brutally slowly: doubling your embryos from 5 to 20 lifts it only 60 percent, because the best of a larger draw is barely better than the best of a small one. Score accuracy is the lever that actually matters, and gains scale with its square root. Here is the full curve at today's IQ-score accuracy (r² = 0.052, Savage and colleagues 2018):
| Embryos screened | Expected IQ gain (points) | What it costs in IVF cycles |
|---|---|---|
| 2 | ~1.9 | 1 |
| 5 (typical cycle) | ~2.5–4.0 | 1 |
| 20 | ~6.4 | ~4 |
| 100 | ~8.6 | ~20 |
| ~500 | ~10 | ~100 |
Read that bottom row twice, because the 10-point gain living in the public imagination as the Gattaca scenario requires screening roughly five hundred embryos. To double the gain at a fixed five embryos, scores would need to explain four times more variance than they do today (r² of 0.21 instead of 0.052), which is beyond any cognitive polygenic score in existence. Score accuracy is the bottleneck, and embryo count is nearly irrelevant past the first handful.
Now the price tag: a US IVF cycle runs $12,000 to $17,000, and MIT Technology Review notes that getting pregnant often takes several cycles. Call it two cycles plus five screened embryos at $2,500 each: roughly $40,000 all in, for a theoretical 2.5 IQ points. That is about $16,000 per IQ point, for a gain that exists in a statistical model and has never been confirmed in a randomized trial of actual children, which means expensive lottery tickets have better-documented payouts.
Here is where the story flips, because the same math that demolishes enhancement quietly endorses something else. Carmi's follow-up paper in eLife (2021) showed that tiny quantitative gains produce large risk reductions once you dichotomize: a 2.5-point IQ shift moves a third of would-be cases across the threshold for intellectual disability, and selecting against high LDL cholesterol with five embryos cuts "high cholesterol" risk by 40 to 60 percent. A 2025 preprint validating nine polygenic scores within families found simulated embryo prioritization delivered 27 to 67 percent relative risk reductions across nine common diseases with just five embryos. As an enhancement tool, embryo scoring is a rounding error. As a disease-avoidance tool, it is legitimately powerful, though you would never guess that from the marketing.
That gap between the pitch and the product has not gone unnoticed. In December 2025, the American Society for Reproductive Medicine's Ethics and Practice Committees concluded that polygenic embryo screening "is not ready for clinical practice and should not be offered as a reproductive service at this time," citing predictive uncertainty and the risk of misleading patients. The European Society of Human Genetics warns there is no evidence the scores can usefully guide one family's choice among its own embryos, since "the genetic variability within an individual family is limited." And the Psychiatric Genomics Consortium, 800 researchers strong, publicly accused Orchid of building its product on consortium data in violation of data-use restrictions. MIT puts the technology on a breakthrough list the same year the fertility field's own professional society says: do not sell this yet.
The strongest case for the other side
It would be easy to end the story as a debunking, and it would be wrong, because the disease-avoidance case deserves its full weight. Parents pursuing IVF are not, for the most part, shopping for IQ points; surveys show their interest centers on reducing disease risk, and only about 11 percent of respondents disapprove of the technology outright. A 27 to 67 percent relative risk reduction across nine diseases, achieved within families with five embryos, is not a rounding error. It is the kind of risk reduction preventive medicine dreams about, delivered before birth. ASRM's verdict targets clinical readiness, not the underlying mathematics, and conflating "not ready to sell" with "doesn't work" is exactly the category error the eLife authors warned against. If embryo screening finds its honest market, it will be as prevention, priced and regulated as prevention, with the enhancement fantasy stripped out of the brochure.
What this analysis does not prove
All of these IQ figures are theoretical. No randomized trial has measured outcomes in children selected by polygenic scores, and Orchid has published no outcome data from screened births. Those scores were built overwhelmingly on European-ancestry genomes, and their accuracy degrades in other ancestries. Reconstructing a whole genome from a few biopsied embryo cells gets genotype calls right 97 to 99 percent of the time at the sites that matter, which sounds high until you remember the signal being extracted is already faint. Prediction intervals are wide: in Carmi's family data, the top-scoring sibling was the tallest in only 7 of 28 families, and one geneticist estimated an intelligence score "might be wrong almost as often as you're right." The r² values fed into these models must be within-family direct effects, stripped of population-structure confounding, and marketing materials rarely disclose which r² they are using, so assume the flattering one.
What you can actually do
If you are doing IVF anyway, monogenic screening where a family history exists remains the highest-value test per dollar; it detects specific devastating mutations with near certainty, which is a different universe from polygenic probabilities. For polygenic add-ons, ask the clinic for the within-family-validated r² behind any trait claim. If they cannot produce one, you are being sold a population statistic applied to your family, which is like being sold the average height of a crowd and told it predicts your child. Watch for two signals that the science has actually moved: published outcome data from screened births (not simulations), and ASRM reversing its December 2025 position. Until then, the regulatory reality is worth understanding: no US federal agency regulates polygenic embryo screening, the European genetics establishment says do not use it, and the American fertility establishment says do not sell it. Right now the market is ahead of the medicine, and the customer is the experiment.
The Bottom Line
Embryo scoring is simultaneously the most overhyped and most underrated technology in reproductive medicine, and the two errors cancel out in public debate. Consider the enhancement story, the one with smarter babies and Gattaca dread: it collapses on contact with arithmetic, 2.5 IQ points per cycle, 500 embryos for ten, $16,000 a point for a theoretical gain. Against it stands the prevention story, the boring one about LDL cholesterol and diabetes risk, which survives the same arithmetic beautifully: up to two-thirds risk reduction with the five embryos a normal cycle already produces. MIT is right to call it a breakthrough. Just not the breakthrough the brochures are selling.