The more autism genes researchers uncover, the more important it is to whittle down the list to those of greatest interest. One productive approach is to map the connections between these genes, building hubs of interactions that may reveal clusters of important players.
Surveying the posters presented Thursday at the 2015 American Society of Human Genetics Annual Meeting in Baltimore, it was clear that — as with all good puzzle-solvers — autism researchers are far from satisfied with their progress. A number of studies involved painstaking refinements to the statistical and systems methods aiming to make sense out of hundreds of autism gene candidates.
In one study, Kevin Lin, a graduate student in Kathryn Roeder’s lab at Carnegie Mellon University in Pittsburgh, presented improvements to DAWN, a statistical method published last year. DAWN builds off of an algorithm called TADA, which rates the significance of an autism gene by looking at how often it has been mutated in people with autism compared with controls.
DAWN adds gene expression information based on the concept of coincidence: Two genes that are expressed at exactly the same time and in the same place might be working together, after all.
So far, DAWN only uses expression information from a single place and time: the prefrontal cortex (at the brain’s surface behind the forehead) during mid-fetal development. A 2013 study identified this brain area and developmental period as critical for autism. But by zeroing in on a single place and time, DAWN draws data from just 107 postmortem samples of brain tissue out of the thousands available in a genetic atlas of the developing human brain called BrainSpan.
“We don’t want to throw away the rest of the dataset just because someone told us it’s not relevant,” Lin says.
Lin scanned data from all the ages and brain regions available in BrainSpan, looking for those that show expression patterns closest to the ones already included in DAWN. “We want to figure out, in a statistical way, which other samples in this entire BrainSpan dataset are related to the ones implicated in autism,” he says.
The original DAWN method found 246 autism genes, of which 18 (or 7 percent) had been independently implicated by another autism sequencing study. The optimized method pinpointed 380 genes, of which 83 (or 23 percent) were listed in the other study.
“We just want a better gene network estimate,” says Lin.
Gene expression patterns are the most common way of connecting autism candidates. But genes are ultimately spun into proteins, which do the actual work in the cell.
In another poster, Jingjing Li, a postdoctoral associate in Michael Snyder’s lab at Stanford University in California, identified proteins that bind to those derived from some of the top autism genes, including ANK2, CHD8 and CUL3.
In a study published last year, the same team looked at where autism candidate genes concentrate within a large network that maps interactions between 13,000 proteins. But that dataset looked for these interactions in kidney cells. Autism genes may be expressed throughout the body, but mutations in them lead primarily to brain symptoms.
“We want to be the first ones to profile neuronal protein complexes,” Li says.
To do this, Li and his colleagues tapped tumor cells from spinal nerves that they transformed into neurons. They then used autism-linked proteins as ‘bait’ to pull out of a neuron any proteins that contact them. They identified their protein ‘prey’ using mass spectrometry.
The genes bound to the autism candidates are more likely to be mutated in people with autism than in controls. They also tend to be expressed during early fetal brain development. What’s more, MeCP2, a protein that influences the expression of other genes, represses the production of many of these interacting proteins. Mutations in the MeCP2 gene lead to Rett syndrome, a developmental disorder related to autism.
For more reports from the 2015 American Society of Human Genetics Annual Meeting, please click here.