Mapping the connections among brain cells could someday prove as revolutionary as mapping the human genome. But tracing synaptic connections between neurons by hand has proven painstakingly slow. Bring on the computers.
Mapping the connections among brain cells could someday prove as revolutionary as mapping the human genome. But tracing each synaptic connection between neurons — essentially a manual effort so far — has proven painstakingly slow. To approach a thorough mapping, researchers will have to develop a computer-automated process.
Even the relatively simple "
wiring diagram" for the tiny
C. elegans worm took more than a dozen years to complete, and that involved just 302 nerve cells. The human brain presents a far greater challenge with about 100 billion neurons, and tens of trillions of synapses that represent millions of miles of wiring between neurons. (Information in the brain travels from one neuron to another across a synapse.)
"In the cerebral cortex, it's believed that one neuron is connected to 10,000 others," said Sebastian Seung, a computational neuroscientist at MIT.
Now Seung is heading a collaborative effort to speed up the mapping of the wiring diagrams, known as
connectomes. He and other researchers want to train computers to imitate human tracing, so that computers can eventually create their own neuron-tracing algorithms and tackle any image of neuronal wiring, no matter how tangled or complex.
Untangling the wires
The main challenge involves analyzing huge numbers of electron microscopic images of brain slices, and tracing the tangled connections that can extend up to several inches between neurons.
One team of neuroscientists at the Max Planck Institute for Medical Research in Heidelberg, Germany, wants to manually trace connections between neurons in the retina, or the light-sensitive tissue at the back of the eye. But as many as 10 people must trace each neuron to catch errors, out of a team of several dozen.
That manual approach would take tens of thousands of work-years to finish the connectome for just one cubic millimeter of brain, according to Viren Jain, a Max Planck Institute neuroscientist who recently finished his Ph.D. under Seung.
Another group has managed to trace the neuronal wiring that connects the brains of mice to the two small muscles that control mice ears. That involved mapping the connections from just 15 neurons branching out to reach 200 target muscle cells, but still involved a "technical tour de force to get all the wires sorted," according to Jeff Lichtman, a neuroscientist at Harvard University in Boston.
"Even though it was a very trivial exercise, it showed us something remarkable and potentially problematic," Lichtman told LiveScience.
Lichtman's success revealed a daunting reality — no single wiring diagram looked the same for any animal. The wiring diagrams for the left and right ear muscles of the same animal also looked different, despite the muscles having an identical purpose. Even a direct comparison of parallel neurons on the left and right side showed completely different branching patterns of connections.
What a brain map can tell us
Researchers have started out with mapping connections among retinas and muscles, because they represent simple challenges compared to the brain. They also know the exact purpose of the neurons and their connections in those cases.
"These things are somewhat easier to understand than if you pick randomly some place in the brain where you don't know where connections are coming from or where they're going, or what they're doing," Lichtman noted.
Neuroscientists still continue to push the boundaries of understanding without having a full wiring diagram of human or animal brains and nervous systems. But Lichtman compared having a connectome to having the human genome mapped out — each a rich data set that scientists can mine for more information.
Having a wiring diagram of the human brain might eventually help answer some fundamental questions in neuroscience, such as how information is organized in the mind. Neuroscientists might also get a better sense of how neuronal connections change over time as people age.
"Where the memory of your grandmother is stored, and in what form it is stored, is almost certainly related to how the brain cells are connected," Lichtman said.
Slicing for science
The National Institutes of Health has launched its own five-year, $30 million Human Connectome Project that starts out simple by aiming to trace the higher-level connections among brain regions, rather than every single connection. Just a few labs around the world have also begun doing their own connectome projects.
That might all change if Seung and his colleagues can truly speed up the mapping with automated computer learning.
"We will be able to test the theory — dating back to the 19th century — that memories are written in connectomes," Seung explained. "We may also be able to find connectopathies, or miswirings of the brain that cause mental disorders."
Lichtman's Harvard lab has already been working with Seung's MIT group on applying new technologies to the task. The researchers have already developed a method of slicing brains thinner than ever before, so that automated microscopes can capture images of the neuronal wiring with unprecedented high resolution.
"Every one of these technological issues is a big challenge, and especially for biologists who are more comfortable with squishy things," Lichtman said.