Annual Report [2018]

The SPARK Gambit

Autism Research (SFARI)

Three years ago, the Simons Foundation Autism Research Initiative (SFARI) took a leap of faith about how best to spur the next generation of gene discovery for autism. The way forward, the initiative decided, would involve letting go of some of the core methodologies of the previous generation of autism gene discovery.

Studies of the initiative’s Simons Simplex Collection (SSC) — a repository of genetic and phenotypic data from families with one child with autism — had uncovered dozens of autism risk genes since its launch in 2006 and indicated that hundreds or perhaps even a thousand different genes underlie the disorder. But these studies had simultaneously made it clear that most of the mutations that cause autism are so rare that they simply can’t be pinned down in a cohort the size of the SSC, which has about 2,600 families.

To identify the majority of autism risk genes would require a much larger cohort, potentially on the order of 50,000 families. But scaling an SSC-type cohort up to 50,000 families would be completely impracticable: The SSC had carried out a meticulous deep dive into each of its families, standardizing diagnostic criteria across its many clinics and bringing each family in not just for diagnosis but also for blood samples, brain imaging and a host of phenotypic measures. Carrying this out with 50,000 families would be prohibitively expensive, and it would be impossible to recruit enough families willing to go through such a time-consuming evaluation process.

“One of the most exciting things for me is that what we’re doing has an immediate impact on families.”

So when SFARI created SPARK (Simons Foundation Powering Autism Research for Knowledge) three years ago with a goal of recruiting 50,000 families, it dropped some of the most ambitious features of the SSC in favor of a more streamlined approach. Families would report their own professional diagnoses of autism and then just fill out online questionnaires and mail in saliva samples. Through this simplified enrollment process, SPARK has already successfully recruited about 18,000 families, and it hopes to hit its 50,000-family target by 2021.

“No one has ever done anything like this at this type of scale,” says Brian O’Roak, a geneticist at Oregon Health and Science University in Portland. “When the SSC got started, it seemed like a phenomenally large number of families, but with SPARK it’s a real paradigm shift.”

But researchers have wondered: Will a cohort assembled in this way provide as clear a window into autism genetics as a collection such as the SSC does? After all, perhaps some of the families who report autism diagnoses to SPARK would not have satisfied the SSC’s rigorous diagnostic criteria. And saliva samples are often contaminated with bacteria, making genetic analyses trickier.

Now, a pilot sequencing study of the exomes — the protein-coding regions of the genome — of 457 SPARK families has suggested, happily, that the leap of faith that went into SPARK’s creation was justified. Genetically, the researchers found, the families in the study seem to mirror previous autism cohorts in a host of ways — the types of mutations that appear, the mutation rate, which genes are affected, and which gene networks and biological pathways are implicated.

“I personally needed reassurance that using this form of recruitment would be worth all the energy we’re putting into it,” says Wendy Chung of Columbia University, SPARK’s principal investigator and SFARI’s director of clinical research. “The good news is, it looks like this is the case.”

“Our philosophy is that as soon as it comes off the machines, we make it available, so that anyone who has a good idea can execute it quickly.”

And as long as participants provided enough saliva, the team found, the quality of the genetic data was just as high as that from blood samples. It was even high enough in the pilot study to allow the researchers to study mosaic mutations — in which only some of a person’s cells are affected by a mutation — which are typically harder to detect than mutations that affect every cell.

The pilot study, published on, was carried out by the SPARK Genomics Consortium, a group that includes members from each of the 25 clinical sites involved in SPARK recruitment.

About 10.4 percent of the families in the study have mutations in one of the approximately 100 genes already known to cause autism. That information has been shared with the families in question. “One of the most exciting things for me is that what we’re doing has an immediate impact on families,” says Tychele Turner, a postdoctoral fellow at the University of Washington who led some of the consortium’s analyses.

The consortium found that approximately 1 percent of the families in the study have deletions or duplications in a chromosomal region called 16p11.2 — roughly the same proportion as in other autism cohorts. And several prominent autism risk genes, such as CHD8, also have mutations in some families.

“I was pleasantly surprised at how great the pilot data were, and how similar this cohort was to the traditional, labor-intensive way of creating cohorts,” says O’Roak, who led the consortium’s study of mosaic mutations.

And the study identified nine new genes that appear to confer risk for autism. Not only do these genes have strong statistical evidence, but an independent systems biology approach showed that the genes have significantly more functional associations with known autism risk genes than chance would predict. 

“Even though we just added 450 families, this already allowed us to discover new genetic risk factors,” O’Roak says. The rate of gene discovery is likely to ramp up significantly in the near future, as the SPARK Genomics Consortium turns its attention to another 9,000 families who were sequenced in 2018. An additional 10,000 individuals with autism (and, where available, both parents) are slated to be sequenced in 2019. A statistical analysis in the pilot study suggests that sequencing 50,000 families is likely to turn up 70 to 75 percent of all autism risk genes.

SPARK made the sequencing data from 2018 immediately available to the broader autism research community, even before the SPARK Genomics Consortium could complete its own analysis of the 9,000 families. 

“Our philosophy is that as soon as it comes off the machines, we make it available, so that anyone who has a good idea can execute it quickly,” Chung says. “Because this is all about powering the research engine of autism to get farther faster.”