By Ashlee Vance

In early October, scientists revealed something spectacular—a complete map of a fruit fly’s brain. Roughly the size of a grain of salt, a fly’s brain has 140,000 neurons connected by almost 500 feet of biological wiring. The map showed myriad types of cells and how they’re connected. It promises to help us gain a better understanding of how brains work. Scientists are rejoicing.

It was a huge achievement, but it also showed how hard it will be to take the next big step forward. Hundreds of scientists spent a decade carving the brain into tiny slices that could be photographed at high resolution, reassembling those millions of images into a unified picture, then analyzing it all with human experts and artificial intelligence software. The previous high-water mark for the field, a whole brain map—known as a connectome—of a worm with 302 neurons, came in the 1980s. Getting to the fly’s 140,000 neurons required a host of breakthroughs.

Next up, researchers want to map a mouse brain, with 70 million neurons. The biggest prize, of course, would be a human brain, in which 100 billion neurons have 100 trillion connections. It’s unclear if it’s even possible to complete such a massive project, or if our computing systems are up to the challenge of making sense of all the data it would yield.

Which brings us to an organization known as E11 Bio.

For the past three years, about a dozen researchers have been working at E11’s laboratory in Alameda, California, to develop techniques to make brain mapping better, cheaper and faster. While the technologies being explored remain in the early stages, Andrew Payne, E11’s chief executive officer and co-founder, sees a path to halving the time needed to map brains while reducing the cost by 100 times.

“In many ways, the brain is the most complicated biological system, and we simply don’t know how it works,” says Payne, speaking about E11’s research publicly for the first time. “This is a big frontier, and there is lots to do.” Like others in the connectomics field, he says better brain mapping could lead to breakthroughs in treatments for diseases as well as improvements in AI systems designed to mimic human brains.

One of the main challenges is making sense of ultradense brain images. Researchers at Harvard University and Google recently produced a three-dimensional image of 1 cubic millimeter of a human brain. It contained about 57,000 cells and 150 million synapses connecting the neurons. The image was built with 1.4 petabytes of data (the equivalent of 14,000 high-definition movies) and shows a chaotic mass of biological wiring not unlike interlocking mounds of spaghetti.

To understand how this hunk of tissue functions, researchers must find a neuron and then trace its connections to other neurons. This is the equivalent of following a single spaghetti strand through a tangle of other noodles to see where it starts and stops, charting all the times it rubs up against the other noodles along the way.

Much of this is done today by AI software, with humans doing the painstaking work of correcting mistakes and picking up on wiring paths the AI missed. The Wellcome Trust, a charitable organization that funds science research, has estimated that it would take more than 15 years to produce a whole mouse connectome and that the process would cost $7.5 billion to $21.7 billion. Most of the cost would come from paying those human proofreaders.

While there are groups funding this type of work in various capacities, it’s not clear that a government or research body would be willing to put up such a staggering sum. “The cost of the proofreading is the biggest bottleneck right now,” Payne says.

E11’s approach, more or less, is to have a brain label itself. The lab does this by injecting a mouse brain with viruses that carry snippets of DNA to individual neurons and instruct the neurons to make special proteins. These proteins then become the biological equivalent of a barcode. If you inject a brain with enough viruses, the individual neurons and other structures end up with combinations of proteins that serve as unique identifiers. The injections go into a living mouse, and it takes about three weeks for the viruses to do their work. After that, the mouse’s brain is dissected.

E11 next sends antibodies into the brain that bring fluorescent dyes to the proteins, so it can use optical microscopes to shine beams of light onto the tissue and pick up the colors. Each neuron lights up with its own color, and its wiring can be traced just by following that color. The brain tissue, in essence, has built-in labels of its various structures and paths of wiring.

Optical microscopes can work with larger tissue samples than electron microscopes, used in most other connectomics research, can. That reduces the amount of brain slicing required. Payne estimates that E11 can analyze an entire mouse brain from a couple hundred slices, instead of the 100,000 slices that would be needed using electron microscopes. In addition, the wiring maps should be more accurate and require less proofreading.

E11 is a focused research organization, or FRO, a concept that Google’s former CEO, Eric Schmidt, helped popularize to accelerate important scientific work. Rather than engage in open-ended research, FROs are meant to hammer away on a specific problem for about five years to see if they can make crucial progress, sharing their findings with the wider scientific community. The theory is that the pursuit of many important breakthroughs is too costly or risky for universities and too commercially uncertain for startups. FROs function as a type of proving ground for novel ideas that could advance science broadly and result in a range of startups.

E11’s investors include Convergent Research, based outside Boston, which has invested about $50 million each in several FROs on behalf of Schmidt and others, and James Fickel, who’s also a major biotech and brain research backer. Payne declines to say how much the lab has raised.

According to Payne, E11’s protein-labeling technology has done well enough for the FRO to shift from proof-of-concept mode to building out its imaging infrastructure so it can tackle an entire mouse brain. Early next year, E11 plans to publish a paper detailing its work mapping part of the hippocampus region of a mouse brain.

Questions have long surrounded the value of these types of brain maps, and they’ve hindered funding for connectome work. Some scientists equate the connectome’s importance to that of the human genome and argue that extraordinary breakthroughs will follow once enough full brain maps are produced. Other scientists say the maps are of limited value, because they’re simply a snapshot of a brain frozen in time and not an insightful window into how the organ operates.

Sebastian Seung, a professor at the Princeton Neuroscience Institute, made major contributions to the fruit fly project and argues that the scientific community is coming around to the value of connectomes. “You have to get to the point where it becomes obvious that the connectome is important for health and disease,” he says. “That’s the way it worked for the genome. I think we are close to that point.”

Seung expects that AI technology will continue to improve and that the costs of brain mapping will come down. It could be the case, for example, that AI solves the proofreading problem on its own. He also points to new brain-mapping techniques being developed at the Institute of Science & Technology Austria and Panluminate Inc., based in New Haven, as possible promising alternatives to E11.

Adam Marblestone, co-founder and CEO of Convergent Research, was drawn to E11 precisely because of the doubts surrounding the connectome field’s ability to keep progressing. Part of the FRO idea is to take risks in compelling areas that have proved too daunting for others to pursue. “There are these different efforts, and there’s potentially going to be more effort with the existing technologies to improve them and go after the full mouse brain,” Marblestone says. “But it’s not like there is a check that has actually been written, because people still wonder if it will cost billions of dollars to do the proofreading. This whole thing depends on dramatically reducing the costs.”