Schizophrenia is one of the most heritable conditions nobody screens for

The parents who ask us about schizophrenia screening have usually been thinking about it for a long time. They watched a father lose a job, a sister stop calling, a son move back home in his twenties. They know what a first break looks like, because they've already been there when it happened to someone they love.

They also know the numbers that apply to their family, even if they've never said them out loud. With one affected parent, a child's lifetime risk is around 10%. With an affected sibling, about the same. With an identical co-twin, closer to a third. Those aren't the population statistics you see in magazine articles. Those are the numbers that actually apply to the family asking the question.

Here's what most of them don't know. Schizophrenia is one of the most heritable conditions in all of medicine, more heritable than breast cancer, more heritable than type 2 diabetes, more heritable than coronary artery disease. The Danish Twin Register, covering 31,524 twin pairs, put the heritability at 79%. And despite that, no IVF clinic in the world routinely offers to screen embryos for it.

The science is there. The profession that would apply it never trained for psychiatric conditions.

A signal this strong, unused

A heritability estimate is the share of variation in who develops a condition that can be chalked up to genetic differences rather than environmental ones. For schizophrenia, that share has been stable across studies and decades. Hilker's Danish cohort landed at 79%. A meta-analysis of twelve twin studies by Sullivan, Kendler, and Neale put it at 81%, with a confidence interval of 73 to 90%. The shared environmental contribution, the part that parenting, neighborhood, and family stress can explain, came in around 11%.

That's not a weak signal. It's among the strongest in clinical medicine.

The genetic architecture is also well-mapped and getting sharper fast. The largest schizophrenia GWAS to date, published in Nature by Trubetskoy and colleagues, combined 76,755 cases with 243,649 controls and identified 287 genome-wide significant loci. Common variants account for about 25% of overall risk, concentrated in genes that govern how neurons form synapses and pass signals. This is a condition built from many small genetic pushes, each one nudging probability up or down, adding up differently in every person.

That's the architecture polygenic embryo screening is built to read.

But the question that matters for IVF isn't whether polygenic scores predict schizophrenia risk across a general population. It's whether they predict differences in risk between embryos that share the same two parents. Embryos from a single IVF cycle are biologically siblings, and siblings are more genetically similar to each other than strangers are. So the validation that counts is the one done on siblings, not on unrelated people from a biobank.

That's the test Moore et al. 2025 ran. Schizophrenia was one of the 17 disease scores checked in the sibling context. Sixteen of seventeen showed no drop in predictive performance within families. The schizophrenia score held up.

What the number on the report actually does

Polygenic embryo screening doesn't tell you which embryo will develop schizophrenia and which won't. No genetic test does that, and any company promising otherwise is selling something the science can't deliver. What it tells you is how the risk variants sort between your embryos, so you can see which ones carry higher or lower genetic predisposition than the others.

Now the part we have to be honest about. The variance polygenic scores explain for schizophrenia is lower than what they explain for metabolic conditions like diabetes or heart disease. In the Moore validation, schizophrenia's liability R² came in around 3.7%. That's a real number, and it's smaller than the 10 to 20% we see for the big metabolic traits. The genetic signal is strong, but the scoring tools are still catching up to the architecture.

That's useful anyway, because the reader of this article is almost never at general-population baseline. A family with one affected parent is already looking at about 10% risk, not 1%. A family with an affected sibling sits in the same neighborhood. At those elevated baselines, even a modest polygenic shift translates to a real absolute change. Capalbo and colleagues raised the concern that a 50% relative risk reduction only moves a population baseline from 1% to 0.5%, which is arithmetically true but uses the wrong denominator. The families reading this aren't at 1%. They're the ones who brought the question in.

There's a second honesty point we owe here. Most polygenic scores were built from European-ancestry data, and psychiatric scores are among the worst affected by that gap. Owen et al. 2023 put variance captured at roughly 8% in Europeans, 7% in Latino populations, 6% in East Asians, and 1.5% in African Americans for schizophrenia. That's a serious calibration problem, which is why we do the extra work to calibrate across eight or more ancestry groups. Screening should work for every family who needs it, regardless of which population was studied first.

The creativity argument, in context

Families who have lived with schizophrenia sometimes hear that screening against it might screen against artistic temperament. The argument comes from a real observation. Relatives of people with schizophrenia are overrepresented in creative professions. There's a modest shared signal with bipolar disorder too, with a cross-disorder genetic correlation of about 0.7, and both conditions have been linked to creative fields.

But the data has a detail that gets flattened in retelling. The relatives show the creative overrepresentation. The people with the clinical disorder don't. A meta-analysis by Acar and colleagues actually found a negative correlation between schizophrenia and creative achievement, r around minus 0.32. Severe psychosis isn't a productive burst of ideas. It's delusions, disorganized thinking, often a decade or more of lost functioning.

The variants that drive risk for the clinical disorder and the variants that underlie creative temperament in unaffected relatives aren't the same genetic signal. Screening reports the probability of the clinical condition. It doesn't screen out personality.

Families who have watched schizophrenia up close already know the difference between a vivid imagination and a full break with reality. They don't need anyone to explain it.

A gap the science can't close on its own

The heritability has been understood for decades. The GWAS has mapped hundreds of loci. The within-family validation has been run. What's missing is a profession trained to bring those results to the families who would use them, and that's the gap Herasight was built to close.

If schizophrenia runs in your family and you're thinking through IVF, our counselors can walk you through what the screening shows for your specific situation. Please reach out to us.