Before virtual brains could see and perform motor tasks as Kate described in the last post, large-scale brain models were being developed in order to identify the synaptic connections between neurons. By mapping this connectome, researchers hoped to be able to explain how information flows through the brain and ultimately examine how perturbations in this information flow leads to neurological disease.
A major group at the forefront of this bottom-up approach to brain modeling is the Blue Brain Project (BBP) in Europe. Since its foundation in 2005, the group of researchers involved in BBP creates this model by recording the activity of an individual cell and imaging its three-dimensional morphology. Each simulated The functional unit of the nervous system, a nerve cell that... requires the equivalent of a laptop computer, making this project very computer demanding. Since they plan to model all 90 billion neurons in the human brain, the BBP has relied on supercomputers to ease the modeling process.
As a proof of principle, the BBP simulated a rat cortical column in 2006. This neuronal network is as small as the tip of a needle and represents a basic unit of the cortex that recurs repeatedly throughout the cortex. A rat’s brain has about 100,000 columns, each containing about 10,000 neurons. Similar types of neurons group together to form 6 overlapping layers of a cortical column. By successfully simulating this network, the BBP showed that they could represent the essential component cerebral mechanics and apply it to a larger scale.
A few months ago, the BBP reported another proof of principle, demonstrating that their model can largely predict the distribution of synapses in the mammalian cortex. The authors looked at the morphology and connectivity of a few neurons from a rat cortical column and formulated general rules of connectivity for different cell types. For example, Cell Type A connects to Cell Type B 90% of the time, and Cell Type C connects to Cell Type B 10% of the time. Then, they put these connectivity rules into the computer simulation, and constraining other known features, like cell density and the ratio of cell types found in each cortical layer. During the simulation, the computer was allowed to randomly form connections between its virtual neurons based upon their connectivity rules. Amazingly, these simple inputs allowed the computer to generate a cortical column that has very similar characteristics to real cortical columns!
It’s remarkable to think that the intricate and complex architecture of the brain may arise from very simple rules and a roll of the dice!Hill S.L., Wang Y., Riachi I., Schurmann F. & Markram H. (2012). PNAS Plus: Statistical connectivity provides a sufficient foundation for specific functional connectivity in neocortical neural microcircuits, Proceedings of the National Academy of Sciences, 109 (42) E2885-E2894. DOI:10.1073/pnas.1202128109 Image adapted from Wikipedia.