An international group of neuroscientists claims to have made a small but significant step toward simulating the entire human brain in a computer. Today, the researchers unveiled one of the most detailed digital reconstructions of brain tissue ever built: a simulation of 30,000 neurons, connected at almost 40 million contact points, in a piece of rat brain about a third of a cubic millimeter in size.
The long-awaited paper is the main outcome of the Blue Brain Project, a 10-year program spearheaded by Henry Markram, a neuroscientist at the Swiss Federal Institute of Technology (EPFL) in Lausanne, and funded in part by the Swiss government. Markram sees it as a validation of plans for the Human Brain Project (HBP), a hugely ambitious project he initiated that aims to model the entire human brain in silico.
But other scientists see the 36-page paper as proof that the idea of modeling a brain and all of its components is misguided and a waste of money. “There is nothing in it that is striking, except that it was a lot of work,” says Zachary Mainen, a neuroscientist at the Champalimaud Centre for the Unknown in Lisbon.
The reaction to the paper mirrors a dispute that has divided Europe’s neuroscience community since HBP was picked by the European Commission as a so-called Flagship project eligible for up to €1 billion in funding. Last year, hundreds of scientists signed an open letter charging that HBP was badly managed and too narrowly focused scientifically. The signatories threatened to boycott the project unless the program was broadened and its governance improved. Two separate, independent reviews largely agreed with the critics, which led to a series of structural reforms at HBP and a diminished role for Markram.
The critics also pointed out that the Blue Brain Project, which started in 2005, had never produced any substantial results. Markram says he feels vindicated by today's publication. “When I started at EPFL, this is what I promised to deliver and I have delivered it,” he says.
The model is the result of an international collaboration of 82 scientists and the culmination of 20 years of experimental and computer work. They set out to model a piece of the sensory cortex where information from the hind limb is processed. In a first step, the scientists had to recreate the anatomy of this area. They took measurements of thousands of neurons in rat brains, from their form and size to the electrical signals they produced. Based on those data and the scientific literature, they came up with 55 distinctive types of neurons. Guided by measurements of cell type density, the researchers distributed thousands of virtual neurons in their simulated brain tissue.
Their virtual brain initially had 200 million synapses—the places where neurons connect and exchange information—which was vastly more than would be expected in such a tiny piece of tissue; the scientists used an algorithm to limit the number to a more realistic 37 million. In a second step, each neuron was matched with a certain pattern of electrical activity, and each synapse assigned to either activate a neighboring neuron or inhibit it. Simulations run on the model reproduced the results from experiments done in cells and animals, the team writes in Cell today.
"As you can imagine, publishing this paper was very tough in a climate of extreme prejudice based on second-hand information,” Markram says. “Now, at least everyone can judge the science directly for themselves.” Moritz Helmstaedter, director at the Max Planck Institute for Brain Research in Frankfurt, Germany, agrees that it's good to have the data out—but the paper confirms his worst suspicions. The Blue Brain Project "has been hyped immensely," Helmstaedter says. "But what happened is exactly what we feared: There are no real findings. Putting together lots and lots of data does not create new science.”
Other scientists also laud the project for the sheer mass of data it has generated while questioning its usefulness. “This was an amazingly huge amount of work, but it teaches us nothing new about how the brain works,” says Peter Latham, a neuroscientist at the University College London.
The model leaves out many features of brain tissue, such as blood vessels and glial cells, which account for 90% of the cells in the brain but are generally not involved in relaying signals. That's because it's only a first draft, Markram emphasizes: “We will include more data in the future.” “That’s like saying, 'I want to go to the moon and I have already put a ladder against this tree,'” Helmstaedter counters.
Another problem is that the model has not been developed with a clear question in mind, argues Chris Eliasmith, a theoretical neuroscientist at the University of Waterloo in Canada. The researchers were successful in replicating certain behaviors, he says. “But you can get all those results with a way less complicated model.” Eliasmith has published a model that includes 2.5 million neurons with about a billion connections, although at a much lower resolution than the Blue Brain model.
Still, the model may be useful in testing hypotheses, Latham says, for instance on how exactly information is relayed in dendrites, the tentaclelike structures leading to a neuron’s main body. An even bigger model may also be more useful, Latham concedes: “I think the Human Brain Project is a waste of money, but after reading this paper I am slightly more in favor of it,” he says. “This paper may actually save the Human Brain Project.”
The Blue Brain team has made their model available online; anyone can examine it on their desktop computer, but to run a simulation you need the power of a supercomputer, Markram says. He says the model has already proved useful for testing hypotheses. For instance, the researchers found that reducing the amount of calcium ions in the model led to a change in the activity patterns, from synchronized bursts of firing to an asynchronous pattern. “There is a transition there. That is an interesting finding,” Markram says—and neuroscientists will be able to uncover others using the model, he argues.
In the end, the question is whether these massive projects are worth the investment, Latham says. “Do you want to spend a billion euros on this? That’s the question.” Markram has no doubts the answer is yes. A hundred years ago, Spanish biologist Santiago Ramón y Cajal made invaluable contributions to neuroscience by peering through a microscope and drawing the cells he saw in brain tissue, he says. “What do you think Ramón y Cajal would be doing today? He would do what we are doing.”