The Miseducation of Autism Heritability

“Study the science of art and the art of science. Learn how to see. Realize that everything connects to everything else,” attributed to Leonardo da Vinci (from How to Think like Leonardo da Vinci, by Michael Gelb, Dell Publishing)

There is an album that did what I once thought impossible, i.e., make my daughter and her friends think I was cool. I discovered Lauryn Hill one night long ago when I stumbled onto her MTV Unplugged performance as I was surfing channels. I stopped just long enough to get hooked. As she sang “I get out—father free me from this bondage”—it was mesmerizing. “I don’t respect your system. I won’t protect your system. I get out.”

Shortly after that, I bought Ms. Lauryn Hill’s “The Miseducation of Lauryn Hill,” her one and only studio album released in 1998. I guess when you make a masterpiece, the smartest thing you can do is walk away (that’s also how I explain not doing the dishes after cooking).

Back to pretending to be cool. It was around 2005, and I was driving my teenage daughter and her friends in a burgundy Honda Minivan down Route 1 in Massachusetts after leaving a Friendly’s restaurant (there was probably a stop at Dunkin’ Donuts at some point too) and I popped in the CD. As “Doo Wop” started to play, I noticed in the mirror that they caught on to what was happening right around the time Lauryn sang “when the cats used to harmonize.” Just then, I heard one of her friends say in a puzzled voice slightly above a whisper: “Your dad listens to Lauryn Hill?” That’s right—soon we were all groovin’ to “Girlfriend let me break it down for you again…”

Anyway, while all the songs on the album are fantastic, the track that stood out the most is actually the only one she didn’t write: a cover of Frankie Valli and the Four Seasons’ “Can’t Take My Eyes off of You.” It’s a brilliant “transdisciplinary” mix of hip hop, doo-wop, pop, and reggae—completely unexpected and completely perfect. It taught me that when you integrate across disciplines and genres, you make something entirely new and far better than the original. Science can learn a lot from the way artists think.

It’s now 2026, and my daughter hasn’t thought I was cool in over 20 years. I’ve been thinking about heritability lately—which admittedly proves her point. Mostly, I’ve been thinking about how we interpret the results.

A Brief History of Heritability

Heritability comes from Francis Galton, a 19^th-century mathematician who wanted to prove that genes drove intelligence. I won’t dwell further on that idea as its silliness tends to rile me up, but he was one of the first to use disease correlation among relatives to study genetics. Ronald Fisher later developed statistical methods to “separate” genetic and environmental variance around 1918, assuming that genes act independently of environment. Keep these dates in mind, as they illustrate how difficult it would have been to study any complex disorder like autism. Twin studies had to be simple enough so that the results could be calculated by hand. Heritability’s assumption that genes and environment act independently came from a time when our knowledge of biology, genetics, and environment was rudimentary. One could argue that Fisher’s assumption grew out of what was possible before computers. This is not meant as a criticism. Assuming genes and environment interacted would have complicated the math in an era in which the abacus was still king. Today, we have AI, genomics, and now exposomics. We can do the complicated math. Today, the reasons we still don’t study gene-environment appear to be an unwillingness to collaborate or to accept reality. To my colleagues in genetics—what harm could possibly be done if we did that?

Despite the fact that studying genomics alone in autism has failed to find its causes or effective new treatments, many insist that this approach must continue because the heritability is high. Heritability has remarkable longevity despite its oversimplification of biology. When actual genetic measures are used in the general population, the heritability plummets from ~80% to ~20%. We then act as though the flaw is in the general population study and not the twin study. We ironically refer to the difference as “missing” heritability, rather than what it really is—unmeasured gene-environment interaction.

Naturally select a result, then look for evidence to bolster our foregone conclusion

Evolution and natural selection also tell us to look at environment over genetics. Gene variants that cause chronic disease will not be selected to be common. Nonetheless, scientific leaders have concluded genetic importance over environment and have been stubbornly trying to find evidence to support that conclusion while disregarding other possibilities.

Which gets me to the point of the blog (finally—right?). The traditional interpretation of heritability is this. If a disease is 70% heritable, then of 100 people with the disease, 70 got it from genetics and 30 from environment. When we speak about the results, this is how we explain them, i.e., there is a nature part and a nurture part. While the calculation is correct, the interpretation is wrong. Can we reinterpret the same findings under the assumption that genes and environment always interact and are always mandatory together to cause a disease? If we assume that this is always true (and it is), what the heritability estimate actually tells us is the relative prevalence and “severity” of both the genetic and environmental factors. In other words, we think 70% heritability looks like the left side of the figure below, but 70% heritability is actually represented by the right side.

All diseases are 100% Genetic and 100% Environmental—including the genetic ones: how to interpret a study showing 70% Heritability

Let’s start by explaining how gene-environment interactions are a mandatory part of all genetic diseases. I’ll start there because if I can show that genetic diseases are always gene-environment interactions, then it is easier to accept that all diseases are gene-environment interactions.

Traditional genetic diseases follow predictable inheritance patterns—like cystic fibrosis or phenylketonuria (PKU). For these diseases, the heritability measured in a twin study will be ~100%. However, that feels like a contradiction. If genes and environment never work independently, why would a twin study ever work? The answer is that heritability is telling us something different than we think. It’s not that genes operate independently of environment—both are always present and working in tandem. However, heritability reflects the relative prevalence and severity of each factor causing the disease. Both are always present. They just don’t vary equally. If the environmental factor is common, the genetic factor will primarily determine if the disease is present.

What I cannot avoid does not make me stronger

All genetic diseases are gene-environment interactions—yes, all of them. Every single genetic disease has at least one environmental factor driving it. What gives the appearance of a “genetic” cause is our ability to “avoid” the environmental factor. For example, cystic fibrosis is caused by a genetic variant impacting the function of an ion channel that impacts chloride, sodium, and potassium transport. Sodium, potassium, and chloride are the environmental components of cystic fibrosis. They are also ubiquitous. We cannot avoid them. The primary cause of the disease will therefore appear to be genetic because the environmental factor is omnipresent and doesn’t vary.

In sickle cell anemia, another genetic disease, the environmental factor is oxygen. In the presence of the genetic mutation, transient low oxygen will cause hemoglobin to form stiff polymers that deform the red blood cells leading to blocked blood vessels and severe pain. We can never avoid intermittent low oxygen in our lives (a mild cold can do it), so the disease appears to be caused by genetics.

Some genetic diseases have environmental factors that are not ubiquitous, just common. With effort, the inciting environmental factor can be avoided. Therefore, we can provide advice on how to avoid complications. Hence, phenylketonuria (PKU) can be avoided if we avoid the amino acid, phenylalanine. One might argue that muscular dystrophy is an example of a genetic disease that doesn’t have an environmental factor involved, although we need muscles to breathe, eat, and drink, so the environmental component is multifaceted and ubiquitous. All genetic diseases have both genetic and environmental components.

Genes never operate independently of environment, but some environmental factors are easier to avoid than others. Try it out. Look up a genetic disease and think through what the environmental factor is. There will always be one. If the environmental factor is impossible to avoid, over centuries, natural selection makes these genetic diseases rarer, not more common.

Everything is Everything

Back to the figure. We’ve been miseducating ourselves on heritability. We interpret it as a genetic effect vs an environmental effect, separating persons with the illness into one of two bins, believing it has to be genes or environment and never considering it is actually both together. We have fallen into the trap of starting with a conclusion and seeking evidence to support that conclusion. Biologically, heritability operates at the individual level—a subtle distinction. It’s never all or nothing—it is always both together. There are no bins.

As shown on the right side of the figure, both are always working in tandem. This explains why heritability can change over time, because the environmental part can change over time. It also explains why diseases can appear to be genetic, even though there is always an environmental component. Note that in the setting of 100% heritability, the two sides of the figure are identical. Finally, this even explains why genetics can be useful to discover treatments, which is the goal geneticists often cite to justify studying the genetics of autism.

The irony of gene-environment interaction

With the exception of gene therapy, all medical treatments are environmental. If a pill, therapy, or another intervention were able to treat autism, that would be proof that environment plays a role. If that pill is theoretically possible, it means that something environmental can affect the gene transcription, translation, modification, regulation, expression, receptor binding, or DNA sequence involved in autism. If a disease were truly impervious to any environmental factors, then treatment would be impossible outside of gene therapy. Having “found” 300+ genes related to autism also proves gene therapy is not a viable option, especially since those 300+ genes explain only a minority of cases. The most relevant question is whether we will find treatments or prevention measures faster by doing genetics alone or by studying gene-environment interactions. We’ve been doing the former for at least 20 years now. It is not working. Remember, we discovered treatments for many diseases, including genetic diseases, long before the age of genomics. PKU is a great example. We have nothing to lose and everything to gain if we modify our course.