New Tool Predicts Autism Risk Genes

A computational technique will help scientists to better identify genes, developmental stages of the brain, and cellular functions implicated in the disorder

brain-specific-network — A gene-interaction network represents how genes in the human brain function together in cellular pathways. Using these network data, researchers were able to predict whether a gene was associated with autism. The network was built by computationally integrating thousands of genomic experiments. Courtesy of Arjun Krishnan/Princeton.

Identifying which genes cause autism is challenging, but a new tool may point researchers in the right direction.

Scientists at the Simons Foundation and Princeton University have developed a computational technique that can sift through more than 25,000 genes in the human genome and rank them based on the probability that they are associated with autism spectrum disorder (ASD). The tool identified hundreds of previously unreported candidate autism risk genes, along with networks, cellular functions and developmental stages of the brain that play a role in ASD. The tool could work for almost any complex disease, the authors say in their Nature Neuroscience paper published online August 1.

“This is the first time that we have a complete list that ranks all the genes in the genome based on their association with autism,” says study co-author Olga Troyanskaya, deputy director for genomics at Simons Center for Data Analysis (to be known, starting in September, as the Flatiron Institute’s Center for Computational Biology) and a genomics and computer science professor at Princeton University. To generate the ranking, Troyanskaya and her colleagues combined a network, or functional map, describing how genes behave in the brain with information on genes previously known to be associated with autism. The researchers then developed an algorithm to analyze network features of the known autism risk genes and used those features to predict the probability that the other genes would be associated with autism.

“Just because we predicted a gene is associated with autism doesn’t mean you say for sure this is an autism gene,” Troyanskaya says. “But you could take the results and sequence the top genes directly to see if they show up in autism patients.” Prioritizing those genes, she says, could help researchers more quickly pinpoint the 400 to 1,000 genes thought to cause ASD. Many of the candidate genes predicted by this approach have been validated by independent sequencing studies — including a Simons Simplex Collection exome-sequencing study — as genuine ASD-associated genes.

In the study, Troyanskaya and her collaborators also combined the list with brain gene-expression data to examine where and when autism risk genes are active during brain development. Changes in the genes associated with autism affect the development of the fetal brain, particularly parts of the cortex that play a role in the control of movements, visual memory, language comprehension, personality expression and decision-making. Activity of autism risk genes, however, happens in regions all across the brain, the analysis showed. The finding reinforces the idea that ASD does not result from one region of the brain malfunctioning but rather from a processing problem that occurs in many regions, the authors write. Using the predicted autism-associated genes and a brain-function map, the researchers also identified specific cellular roles that underlie the association of those genes with autism, even for genes whose functions were previously not known.

Troyanskaya and her colleagues also used their autism-risk-gene ranking to examine autism-associated copy number variants (CNVs). CNVs are regions of the genome that are duplicated or deleted. Previous studies have not been able to pinpoint which genes in these regions cause autism, but the new ranking identifies the ones that are most likely at the root of the condition. The analysis also showed how genes from different CNV regions converge and how their relationship may disrupt cells and cause autism.

In addition to helping to direct future sequencing studies, the results will help biomedical researchers hone in on novel autism risk genes and guide their studies on the developmental and functional effects of these genes. All data from the study are freely available at http://asd.princeton.edu.