Science Editor

Like so many neurological disorders, autism clearly has a genetic component. But the details of those genetics have proven very hard to pin down, partly because – as, indeed, the name "autism spectrum disorder" makes obvious – individuals with a wide spectrum of symptoms ultimately funnel into a diagnosis of autism.

Such variability has only increased over time, as the criteria for diagnosing autism have become more broad, and high-functioning children whose only label would once have been "quirky" more often receive a formal diagnosis.

By taking this variability into account, researchers from George Washington University have managed to identify 18 single nucleotide polymorphisms, or SNPs, that are markers for autism spectrum disorder.

The researchers were able to identify those SNPs by first separating individuals with a diagnosis of autism spectrum disorders into separate subcategories according to their symptoms, and then analyzing each subcategory separately.

"Most geneticists think that if you have small numbers, your [statistical] power is going to be lower," Valerie Hu told BioWorld Today. "And yet, they also acknowledge that there is enormous heterogeneity within the autistic population."

Hu is professor of biochemistry and molecular biology at George Washington University's School of Medicine and Health Sciences and the study's corresponding author.

The work was published in the April 27, 2011, issue of PLoS ONE.

The studies were based on earlier work by Hu and her colleagues doing so-called quantitative trait analysis, which consists of scoring autistic children on more than 100 categories in five separate traits such as language development and social skills.

Based on quantitative trait analysis, the researchers sorted the nearly 1,900 subjects into four separate groups of between 350 and 650 individuals each, and compared the relative frequencies of about half a million SNPs between the subgroups, the whole cohort and roughly 2,500 controls.

In their paper, Hu's team described those four groups as "one with severe language impairment, another with mild severity across all items, a third of intermediate severity, and a fourth of moderate severity with a higher frequency of savant skills."

In initial analyses, the authors identified 167 SNPs that they further analyzed within each of the subgroups. Of those 167, eight were associated with autism in one of the subgroups, and an additional 10 were associated with autism spectrum disorder in more than one subgroup.

"Some of the profiles overlapped – as one would expect, since they are all autistic. But others were unique to the subtypes," Hu said.

Many of the SNPs were in regions that have been identified, in other studies, as regions where autistic individuals had copy number variations, further validating the findings. Strikingly, none of the 18 SNPs were in the protein-coding regions of the genome.

Hu said that this suggests they have regulatory functions that might affect many proteins, rather than changing the structure of one specific protein, which might explain why autism symptoms are pervasive. "If they have regulatory functions, then they can really have a much greater impact than if they were just affecting one gene."

But what those regulatory effects might be is still a mystery.

"We know so relatively little about the functions of noncoding DNA," Hu said.

Several of the "p" values had a "p" value that was greater than the cutoff, having instead what Hu and her team termed "suggestive significance" of less than 0.1, but greater than the classical cutoff of 0.05.

Hu's team acknowledged in the paper that different researchers use different statistical methods in determining whether a SNP is significant, and that some might choose more stringent criteria, and Hu acknowledged the same thing when speaking to BioWorld Today about her data. "I'm not saying that we have developed the best way," she said. "I'm saying that we have developed one way that seems to work."

And that method, she added, might prove useful for discovering genetic markers of other diseases whose clinical symptoms vary widely, in particular other neuropsychiatric disorders. "A lot of the difficulty in many of the neuropsychiatric disorders," she said, "is the heterogeneity."