News The latest developments in autism research.
Profiles Portraits of scientists who are making a mark on autism research.
Toolbox Emerging tools and techniques that may advance autism research.
Spotted A roundup of autism papers and media mentions you may have missed.
Opinion Conversations on the science of autism research.
Viewpoint Expert opinions on trends and controversies in autism research.
Columnists Dispatches from experts on various facets of autism.
Crosstalk Debates and conversations about timely topics in autism.
Reviews Exploring the intersection of autism and the arts.
Q&A Conversations with experts about noteworthy topics in autism.
Deep Dive In-depth analysis of important topics in autism.
Special Reports Curated collections of articles on special topics in autism.
Webinars Presentations by leading experts on their latest research.
Opinion / Cross Talk

Should autism research focus on common or rare risk factors?

by  /  20 August 2014

In July, we reported on a large population-based study, published in Nature Genetics, which estimates that common genetic variants compose about half the risk for autism.

The study analyzed new epidemiological data from a large Swedish sample, as well as from the Autism Genome Project, the Simons Simplex Collection(funded by the Simons Foundation,’s parent organization) and other groups.

Some geneticists found the estimated contribution of rare variants — at about 3 percent — unexpectedly low. The work also raised questions about whatresearchers should do with these ‘pies of risk.’

Should researchers focus more on common variants, rare variants or spontaneous non-inherited ones? How should researchers weigh data from studies that model proportional risk across populations, versus risk in individuals, for whom rare mutations with large effects are more significant?

We sought answers from leading statistical and analytic geneticists who study risk factors for neuropsychiatric disorders, as part of our discussion seriesCross Talk.

What do you think? Share your reactions and follow-up questions in the comments section below.

The Experts:

Elliott Sherr

Professor of Neurology, Pediatrics, University of California, San Francisco

Michael Owen

Director, Neuroscience and Mental Health Research Institute

Benjamin Neale

Assistant Professor, Instructor, Associated Researcher, Massachusetts General Hospital, Harvard Medical School, Broad Institute

An underestimate of non-inherited burden

Professor of Neurology; Director, Comprehensive Center for Brain Development, University of California, San Francisco

Surprising slices: “The Nature Genetics paper is a bold attempt to model the causes of autism. The results suggest that the heritability for autism is high — in the range of 50 to 60 percent — but that most of this heritability, according to their model, comes from common and not rare variants. Moreover, the risk of acquiring autism from de novo copy number variations — spontaneous duplications or deletions of large regions of DNA — or from de novo single nucleotide variationseems unexpectedly low in the general autism population.”

Beyond the exome: “The study provides a novel and exciting framework for considering autism risk. However, there are a few limitations worth considering. One concern I have is the way they incorporate de novo mutations into their model. The authors only examined de novo single nucleotide variants that are either missense or truncating, and they captured only a subset of noncoding regions. This approach certainly leads to an underestimate of the overall burden of de novo variants in autism, especially of noncoding mutations. There were no apparent adjustments in the model for variants that lie outside of the exome, the protein-coding region of the genome.”

Autism higher in dads: “Studies over the past five or so years show a higher rate of de novo mutations per generation when looking at the sperm of aging fathers. Those results suggest that a significant percentage of the overall autism risk for populations arises from this higher de novo rate. The authors of the Nature Genetics paper, however, do not account for this in their model.”

Diagnostic domains: “Another issue for any large population-based study relates to the issue of diagnosis. While the researchers here do deal with the difference between autism, autism spectrum disorder and other broader definitions, they do not address the deeper biological question of what ‘unit’ of autism is truly heritable within families. For example, if we were to refine how we measure the different domains of social impairment, we might find much higher heritability if we separately measured the degree of social motivation or social anxiety within families, instead of just measuring the overall autism with a thumbs-up or thumbs-down approach to diagnosis.”

Common and rare paths forward

Director, MRC Centre for Neuropsychiatric Genetics and Genomics; Director, Neuroscience and Mental Health ResearchInstitute, Cardiff University, U.K.

Common interests: “These findings are of interest for a number of reasons. First, they show that heritability estimates — whether based on twin studies or newer sequencing techniques — are converging around 50 percent. Such estimates carry a degree of uncertainty, but they do suggest that environmental and/or random variation as well as genes are implicated in the disorder. Second, the findings indicate that common genetic variants play a substantial role in autism.”

Go big: “This is not surprising. The findings show that both common and rare variants play a role in autism and that, in this regard, it is similar to other common disorders, notably schizophrenia. Comparison of findings from autism and schizophrenia suggest that the failure of geneticists to identify specific common risk variants is likely due to the fact that autism researchers simply haven’t studied large enough samples.”

Rare or common? “The question now is whether to focus on common or rare variant discovery, or both, and this comes down to what will give us the best handle on the neurobiology and identifying new treatment targets. Identifiable rare risk alleles tend to have relatively large individual effect sizes. In contrast, common alleles are associated with low individual risk and are less readily actionable in neurobiological studies.

“However, given that common variants can now be seen to contribute quite substantially to autism risk, it would be sensible to push on and identify specific common associated variants in the hope that, as with other disorders, this will implicate specific areas of biology that impact a substantial proportion of cases.

“The way forward should include both common and rare variant studies. In fact, both will depend upon access to very large samples, and the autismcommunity needs to now focus on putting together many tens of thousands of genomes to study.”

A robust genetic starting point

Assistant Professor of Analytic and Translational Genetics, Massachusetts General Hospital

Promising trajectory: “Over the past few years, there has been a serious debate in the community about the role ofcommon variation in neuropsychiatric disorders such as schizophrenia and autism. Some of the earliest genome-wide association studies (GWAS) for schizophrenia demonstrated that there was polygenic inheritance — many small-effect variants scattered across the genome. But these early studies failed to identify significant loci. With newer GWAS of schizophrenia, however, we have found more than 100 significantly associated loci, effectively removing all doubt as to whether a genome-wide approach can identify regions relevant to disease.

“GWAS of autism are in a similar position as the early attempts in schizophrenia, failing to identify unequivocal genome-wide significant loci. However, the work presented here strengthens the case that they may in fact yield significant loci for autism.

Complicated numbers: “In terms of the precise estimates and how valuable they are, the story is a little more complex. Any attempt to quantify what fraction of the variance in a trait is attributable to common genetic variants requires a number of assumptions as well as considerations in interpretation. For example, the estimate in this paper is much higher than what was seen for the Psychiatric Genomics Consortium analysis using a similar methodology, but both datasets support an important role for common variation.”

Incomplete story: I think the landscape is quite complicated and that chasing both common and rare variants has value. It is most certainly the case that analysis of non-inherited mutations in autism has successfully identified novel risk genes and variants. That is without question at this point. However, strong mutations with large effects may not tell the whole story. If we consider other complex traits, such as lipid levels or height, it is certainly the case that there are strong-acting variants that can cause individuals to have extremely high or extremely low lipids or be extremely tall or extremely short. The GWAS of these traits, however, reveal a much more complex biological landscape than can be pieced together simply from powerful mutations.”

About Cross Talk
Discussions among leading experts in the field. Submit your questions to [email protected].
  • naomiwray

    Readers of the Gaugler et al study discussed here should not get bogged down in decimal point comparisons of estimates of heritability, but should see the bigger picture that the benchmarking provided by heritability estimates provides strong pointers for what to prioritize in the next phase of autism genetics research. Their analyses suggest that common variants associated with autism exist and are worth identifying as part of the portfolio of strategies to increase understanding of the causes of autism spectrum disorders.

    This conclusion may seem surprising to some, since enrichment for de novo mutations and rare copy number variants are the only discoveries in autism genetics research to date. Gaugler et al do not undermine these important results, but quantify that they explain only a small fraction of variation in risk and an even smaller proportion of the heritable contribution to risk. Genome-wide association studies (GWAS) have been successful for many complex genetic diseases, but so far they have not delivered any genome-wide significant associations for autism. However, Gaugler et al use currently available GWAS data to estimate that common SNPs explain an important contribution to variance. This provides empirical evidence that current GWAS are underpowered and that individually associated common risk variants will be identified as sample sizes increase. These results mirror those for schizophrenia: predictions that increased sample sizes would deliver genome-wide significant associations have been vindicated, with > 100 common risk alleles now identified for schizophrenia.

    Individually, common risk variants will have small effect but together are important for explaining population risk. Identification of a mass of risk variants will give pointers to risk pathways. The next challenge is collection of very large phenotype-rich cohorts. Genotyping technology is now so cheap that the funding needed for genotyping is trivial compared to the costs of cohort collection. This strategy will give biggest bang for buck in the short-term and will position well for future studies designed to identify less common risk variants.

    • RAJensen

      The rate of de novo mutations are vastly under reportd in this study Here is the exclunionary criteria this study used in identifying their final sample:

      ‘Exclusion criteria were individuals with a diagnosis of autism, but who also had a genetic disorder known to be associated with autistic features (e.g., Fragile X, Down and Klinefelter syndromes); or medical or psychiatric history that could mitigate a confident diagnosis of autism’.

      Genetic syndromes account for a large sample of the ASd population and most genetic syndromes are caused by de npvo mutations arising in reproductive cells.

  • RA Jensen

    Autism is a multifactorial condition that in most cases likely involves multiple risk factors of small or moderate effect that cumulatively increases total individual risk. A great deal of argument about common genetic variations accounting for 60% of autism risk is based on a widely distributed recent paper published in Nature Genetics. How common are heritable genetic variations associated with developmental problems including autism? Turns out the liability to autism associated with common genetic polymorphisms are quite weak but not entirely meaningless.
    Catechol-O-methyltransferase (COMT) Gene (22q11.2). The deletion of the COMT gene produces the 22q11 deletion syndrome with relatively high autism risk. Common polymorphisms in the COMT gene have also been associated with autism risk. How common are these polymorphisms? The Val108/158Met (rs4680) polymorphism has been associated with autism risk. Research in Mexico found the prevalence of this polymorphism ranges from 57% (Tabasco) to 85% (Mexico City). The authors analyzed the frequency of Val108/158Met polymorphism of Caucasian (54% Met allele), Asian (29% Met allele) and African (34% Met allele) populations separately and also in comparison with Mexican (63% Met allele) population.
    Chromosome 7 is the largest chromosome and mutations in chromosome 7 were the first identified autism candidate genes. The MET gene (7q31) has been reported as a candidate autism gene. Three groups have reported that the rs1858830 C allele variant in the promoter region of the MET gene on chromosome 7q31 is associated with autism. The rs1858830 C allele variant, is associated with autism, neuronal development and gastrointestinal disorder, however the variant is present in 47% of the general population which is a non-significant risk factor of weak effect.

  • RA Jensen

    The serotonin transporter gene, SLC6A4 (Chromosome 17q.11-12) was first found to be associated with autism in the 1990’s. In the largest study published the short S allele in SLC6A4 was found to be over transmitted in autism.

    How common is the short allele in the general population?
    In the human population, the frequency of the long allele of 5-HTT is approximately 57%, while that of the short allele is 43%.

  • RAJensen

    Finally here is another study that doesn’t quite fit the UK and US ‘autism’ is the most heritable of developmental disorders’ claim promulgated by child psychiatrists and molecular geneticists. Genetic influences do operate in Hansen’s disease (Leprosy). Like autism, leprosy is multifactorial involving genetic influences (susceptibility factors) but is always caused by exposure to the mycobacterium leprae pathogen. A study of families in a small island with a high leprosy infected population was published. A small study but eye-opening findings were reported. They calculated a heritability estimate of 57% similar to many autism twin studies. The recurrence risk for siblings under 21 was 6.4%, similar to the autism recurrence risk seen in Scandinavian sibling recurrence risk estimates. Is it time to drop ‘heritability’ from the autism scientific jargon promoted by child psychiatrists and molecular geneticists?


Log in to your Spectrum Wiki account

Email Address:



Request your Spectrum Wiki account

Spectrum Wiki is a community of researchers affiliated with an academic or research institutions. To be considered for participation, please fill out this form and a member of our team will respond to your request.


Email Address:

Title and Lab:

Area of Expertise: