Crowdsourcing Study of 30,000 Images Connects Genes to Brain Size
DNA data and medical images were combined to find genetic links to brain anatomy.
The brain remains an enigma to geneticists, but big data strategies could help.
A large network of neuroscientists and doctors that compared over 30,000 brain images with people’s DNA says it’s found several genes that appear to influence the size of brain structures involved in intelligence and memory, as well as the volume of the brain itself.
Although the medical importance of these clues remains far from clear, the consortium, called Enigma, says its work demonstrates a novel distributed-computing strategy able to sort through vast numbers of MRI scans and DNA test results. “What Enigma is doing is combing through every pixel of every scan and comparing it to every genome,” says Paul Thompson, the neuroscientist who organized the research. “This is a roadmap to how you do this.”
Thompson, who is head of the Imaging Genetics Center at the University of Southern California, believes Enigma is the largest collaboration ever to combine efforts to study the brain. The new study, published today in the journal Nature, lists 287 authors and 193 institutions. The study involved the analysis of 30,717 brain scans as well as DNA information gathered by researchers in Cambodia, South Africa, the United States, and other countries.
MRI scans are expensive, and analyzing them requires intensive computation. Especially when combined with DNA information, these data are too large to easily move over the Internet, and in some cases privacy rules prevent them from crossing national borders. The consortium says it is solving this problem with a distributed approach in which all the centers are given common algorithms with which to process their images, after which then their findings were weighted and combined.
Large studies linking genes with disease have not paid off well in neuroscience. For some common conditions, like depression, there are no convincing DNA clues at all. Instead of giving up, however, some researchers are instead seeking ways to vastly increase the size of studies. In the case of Enigma, that happened by “crowdsourcing” analysis of existing MRI scans. “There are brain images of many kinds on Alzheimer’s, schizophrenia, and autism that people have collected for decades. There is just astronomical data that is siloed,” says Thompson.
The big data approach is in vogue. Last year, the National Institutes of Health awarded Enigma and several other centers $32 million as part of a plan by the funding agency to plow more than half a billion into new ways of exploiting biological data over the next seven years. In an interview last month, Mark Guyer, an adviser to the NIH program, called Big Data to Knowledge, says the agency believed data analysis, not collecting data, was now the “bottleneck” to research.
Enigma capitalized on what would be more than $30 million worth of existing brain scans (assuming each cost $1,000), taken of people who ranged in age from age nine to age 96. By comparing the scans to people’s DNA, the Enigma researchers say they found eight regions of the genome that influenced either the overall size of the brain, or the volume of its substructures.
The strongest effects were found in the putamen, a part of the brain that influences learning and movements, and which is notably smaller in people with Parkinson’s and Huntington’s, says Thompson. In a person with two of the gene variants identified, the structure would be 2.8 percent smaller, according to the research.
Despite the scope of the effort, the size of brain structures weren’t successfully linked to any psychiatric illness, although Thompson says there are clues pointing in that direction. “It might not be as simple as the gene gives you a smaller putamen and you get these diseases, but the genes are likely to affect how many cells you have, and how they get to the right place,” he says. “It’s vital to know how it’s built.”
To critics, the Enigma project embodies the shortcomings of big, mathematics-driven biology, which focuses on data that are easier to feed into computers. Evan Charney, an associate professor at the Duke Institute for Brain Sciences and its public policy school, called the study “depressing” because of how it downplayed life events and environmental influences, like exercise and stress, that also affect the brain’s anatomy. “None of this played any role in the authors’ analysis,” he says.
Thompson says he is convinced that more data, and new mathematical techniques, will eventually lead to substantial breakthroughs in brain science as they have in other domains, like speech processing or the cracking of the WWII-era Nazi cipher after which the consortium is named. He says until recently it was “heresy” to even suggest that gene variants could be linked to what’s seen in medical images. “People said you will never see the effects of personal genomes in brain scans,” says Thompson.
Become an MIT Technology Review Insider for in-depth analysis and unparalleled perspective.Subscribe today