By Grant Jacobs 24/10/2017 1

Ever since Rain Man, autism has been a condition du-jour for those touting (fake) ’wellness’ remedies, opposing vaccines, and general head-nodding concern. Some claim autism is caused by any number of things, offering just many remedies offered to “treat” it. Recent research reminds us that autism is mostly genetic.

People with autism, or autism spectrum disorder, have difficulty in social communication and interaction. One autistic computer programmer wrote to me that he preferred to interact by email, rather than meet with me in person. It can also feature repetitive behaviours such as stacking the cans in the feature image for this article, ordering toys in rows, and strong interests on particular narrow topics.

Hollywood movies like Rain Man and TV dramas are more inclined to idolise ‘autistic savants’. Curiously Kim Peek, the real-life inspiration for the Rain Man character played by Dustin Hoffman, was not autistic. Brain scans showed that he had corpus callosum agenesis as described towards the end of my earlier article, The mutant PRICKLE and the split brain.

For a more real-to-life setting, a film outlining features of autism in young kids can be found at the ASDinfoWales website. One excellent summary of autism myths and facts, with research references, can be found at the UK Austistic Society’s Autism facts and history page.

Placing toys in ordered rows can be important for some autistic kids. (Source: Wikipedia, GNU Free Documentation License. Edited to preserve anonymity of the subject owing to model license concern.)

There is a long history of autism being considered mostly genetic; autism is often said to be the “most genetic” of the neurodevelopmental disorders. For example, twin studies point at high rates of heritability, around 90%. Similarly, siblings are 15-30 times more likely to have autism than the general population.

It’s also increasingly thought to start very early, before the chid is born. Autism is more frequently found in boys, and there are associations with parental age and other factors.

If you were to peek into the research publications estimating the extent to which autism is inherited or genetically-determined—two slightly different things!—you’d find a thicket of numbers and arguments, with values from around 60% to 90%, with a few outliers.

It’s such a jungle, you’d want to go in armed with a machete in each hand to cut your way through the mess. (And toss on a bandolier or two, and a bunch of automatic weapons, Indiana Jones style just to look the part.)

One study latched on by some of those touting “remedies” for autism is a study suggesting a hereditariness of about 50%. Researchers made some suggestions why this study offered a much lower estimate of hereditary contribution to autism than other estimates. A fresh re-examination of the dataset comparing different methods shows the variation in autism cases is mostly genetic, at around 83%.

These statistics are often incorrectly reported as saying something about each autistic person. It actually tells us about what fraction of the variation in autistic people as a large group is due to inheritance. There’s also one type of genetic contribution that is not inherited that is not considered in hereditary estimates: “brand new” (de novo) mutations.

The researchers used same dataset they previously used—over 2 million Swedish children—this time applying a method “used by previous studies in the field”, finding the heritability to be 83%, rather than the 50% they previously found.

Clearly the method used to estimate heritability matters. In their earlier work, they used a mathematical model that included “time to ascertainment”: when it was that the child was diagnosed. One concern was that it might the hereditariness to be underestimated. (Siblings are often diagnosed at quite different times, and sometimes under diagnosed.) The re-examination of the data suggests this is what has happened.

Re-examination of assumptions of previous work is part of how science proceeds. It must seem a slow, one step back and a step or two forwards process to the public, but that’s how it goes.

It also conflicts with the tidy stories media want to tell. If you were to go out and read articles about this recent estimate, you’d find very little pointing to the wider, messier story I alluded to earlier.

These estimates are quite complex. They also rely on a bunch of assumptions. I suspect that it will be difficult, possibly not even practical, to try obtain a “correct” figure beyond the broad observation that autism is strongly genetic.

Beyond diagnostic and sampling issues, the genetics of autism itself provides challenges. There are small number of “simple” (Mendelian) genetic mutations that might explain perhaps 10% or so of autism cases. For these few cases, an association between autism and a well-defined mutation that is inherited in a simple way has been found. There is also evidence that a very large number of genes is involved, each contributing to a small fraction of the cases.

Furthermore, there is the issue of mutations that have arisen anew in the child. (I’ve previously written on one study of these in, Autism genetics, how do you copy?) Each of us has perhaps a hundred or so mutations not present in either of our parents. Most of these mutations do nothing, and have no effect on us, but some types of these so-called ‘de novo’ mutations are associated with different conditions.

There tend to be more of these de novo mutations in children of older parents, and having older parents has been associated with autism. For example, sperm get slightly more mutations in them as the father ages. As they’re not inherited from the parents, these de novo mutations don’t add to the “hereditary” side of the tally. They’re considered to be “environmental” even though they’re also clearly a genetic thing! (The trick is to notice the difference between ‘inherited’ and ‘genetic’: they’re genetic, but not inherited.)

The complexity gets worse as you look further!

One thing that might be useful is more information on physical things we can record and observe in genome data. That would require very large collections of genome data, accompanied by good medical records. There are some attempts to look at aspects of this. They each look at one type of genetic change, in subsets of the genome data. Essentially they’re taking educated guesses at what might be useful to look at. It’s a reflection of the technology and numbers involved. Simply surveying the genomes outright, as whole genomes, will involve even more genomes than the current large-scale genome sequencing efforts are producing. (And a lot of money!) As one example GenomicsEngland report they have sequenced just over human 36,000 genomes as I write. Compare that to the two million-plus children used in the recent estimate of hereditariness.

It might, then, be a case of waiting for the technology and costs to become suitable — which they will, and surprisingly soon: years, rather than decades.

Other articles on Code for life

The mutant PRICKLE and the split brain

Tracking disease and human migration through genetics

Preserving endangered species – of gut microbes

Autistic children and blood mercury levels

Are too many vaccines too soon harmful?

Visual illusions, change blindness and autism

Autism genetics, how do you copy?


JAMA. 2017 Sep 26;318(12):1182-1184.

The Heritability of Autism Spectrum Disorder.

Sandin S, Lichtenstein P, Kuja-Halkola R, Hultman C, Larsson H, Reichenberg A.

PMID: 28973605 DOI: 10.1001/jama.2017.12141


My thanks to a colleague for supplying me with a copy of the paper.

The free personal registration for JAMA is limited those in the USA. I’ve no idea if this would let me get a full copy of the research paper if I faked being from the USA with VPN!

Featured image

Source: Wikipedia, used under GNU Free Documentation License. “Quinn, an ~18 month old boy with autism, obsessively stacking cans. Date: Late 2002. Place: Walnut Creek, California. Photographer: Andwhatsnext. Scanned photograph. Credit: Copyright (c) 2003 by Nancy J Price (aka Mom)”

One Response to “Autism is mostly genetic again”

  • Adding as a quick comment in case I don’t get to writing about it elsewhere (fairly likely these days!) –

    Developmental variation is another thing to consider. This variation isn’t considered “genetic” in these calculations, but they’re not really “environmental” in the that usually means. The terms used for this work are historic, and in some ways a bit of a nuisance.

    This variation is epigenetic, in the original Waddington sense of the term, not the much hyped epigenetic inheritance, let along the truly strange “epigenetics” from some people who tout unscientific “remedies” for any number of things (or Deepak Chopra’s idea of epigenetic, either!)

    As always, there is far too much I could write about…