How Do You Assemble a Brain? Randomly

It is a puzzlement: How do you assemble and wire an information processing device as complex as the mammalian brain? There are roughly 86 billion neurons in a human brain, forming about a quadrillion synapses. A rat’s brain is just one thousandth that size, but still pretty complex, with 56 million neurons and 500 billion synapses.

How does the brain know to put a nest basket cell here, a small basket cell over there, a large basket cell in the middle, a Martinotti cell on the left and a bi-tufted cell on the right, all wired up to pyramidal cells? There has to be a plan, doesn’t there? I mean, the body doesn’t just throw its inventory of brain cells out there like a bunch of pick-up sticks, to fall where they may.

As it turns out, that may be almost exactly what the brain does. Like so much else in capital-L Life, connections in the brain may be emergent: the developing brain lays out its thinking cells in a nearly random mixture, and then wires them up after the fact.

The Blue Brain group (motto: “Reconstructing the brain piece by piece and building a virtual brain on a supercomputer”) at Switzerland’s Ecole Polytechnique Federale de Lausanne (EPFL) has built a computer model of a 298-cell slice of rat cerebral cortex. The model distributed the 6 types of neurons randomly, according to their frequency in natural tissue. They tracked “the incidental overlap of axonal and dendritic arbors,” the tree-like branchings at either end of the nerve cell that reach out and form synapses.

It’s a case of, “Ready. Fire. Aim.” The cells are (mostly) laid out, and the synapses appear to grow when the right cells touch one another—with some small role perhaps played by chemical or other signals that cause a few cells to reach out to one another after the fact.

Other studies have already established that random distribution could produce the proper number of potential neuron-neuron connections. The EPFL study (published online by the Proceedings of the National Academy of Sciences on 18 September) is first to demonstrate that chance could account for the proper types of connections—the right number of connections between nest basket cells and pyramidal cells, and vice versa.

The researchers used their software tools (BlueBuilder) to model the nervous system. They then reduced the synapse-identification problem to a very large number of computer graphic “cylinder to cylinder touch” assessments, running on a 16 384-cpu IBM BlueGene/P computer at the Center for Advanced Modeling Science (CADMOS) in Lausanne.

Finally, they compared the results to data from cross-sections of actual rat brains, stained and digitized, and analyzed to show cell types and synapse locations, and found an overall 75% correspondence between the two. Not absolute correspondence…but a good deal better than chance.

Images: EPFL / Blue Brain Project