The Ultimate Connection Machine
Tilikum the killer whale (Orcinus orca) made news recently in the tragic death of his Sea World trainer, Dawn Brancheau. Tilikum pulled Brancheau into the water when he grabbed her floating ponytail — much like a cat might grab yarn attached to a stick. Complex play behavior is a sign of intelligence, but unfortunately little is known of the circuitry of even a cat’s brain, much less the massive brain of an orca — roughly four times the size of a human brain.
MIT neuroscientists are developing computerized techniques to map the millions of miles of neuronal circuits in the brain that may one day shed some light on the differences between Homo sapiens sapiens and other species, and will likely clarify how those neurons give rise to intelligence, personality, and memory. Developing connectomes (maps of neurons and synapses) may have just as much impact as sequencing the human genome. Here’s a video showing 3D rotating nodes and edges in a small connectome:
Just as we can now compare individuals’ genes to look for differences that might account for mental illness, cancer, or other diseases, neuroscientists will be able to use connectomes to help identify diseases such as Alzheimer’s and schizophrenia. The connectome wiki was established as a knowledge base for macro and meso-scale brain region and brain connectivity information across species. Eventually we may be able to compare the wiring of a species like Felis silvestris catus (the domestic cat) — or even the orca — to the wiring of our own brains.
The human brain has an estimated 100 billion (1011) neurons. In comparison, the original “massively parallel” CM-1 connection machine built by Danny Hillis in the 1980s — based on ideas from Nobel prize-winning physicist Richard Feynman — had just 65,536 individual processors. Today’s fastest supercomputer (the AMD Opteron-based Cray XT5 Jaguar at the Oak Ridge National Laboratory) has 224,256 cores and a sustained processing rate of 1.759 PFLOPS (one thousand trillion floating point operations per second). The human brain, in comparison, is estimated to have a processing rate of approximately 10 PFLOPS.
How do researchers intend to map something so complex and massively parallel as the human brain with its billions upon billions of connections? The first connectome was mapped in the 1970s by researchers at Cambridge University for C. elegans, a tiny worm about a millimeter long. It took more than 12 years. Traditional approaches to connectome mapping include optical microscopy for cell staining with the injection of labeling agents or reconstruction of tissue blocks using electron microscopy. Researchers at MIT’s Seung Lab want to speed this up by teaching supercomputers to learn by example. “They [MIT neuroscientists] feed their computer electron micrographs as well as human tracings of these images,” reports MIT News. “The computer then searches for an algorithm that allows it to imitate human performance.”
The MIT research is part of a larger effort known as the Human Connectome Project (HCP) funded by The National Institute of Health’s Blueprint for Neuroscience Research. According to NIH’s press release announcing the project, this $30 million effort will use cutting-edge brain imaging technologies to map the circuitry of the healthy adult human brain. By systematically collecting brain imaging data from hundreds of subjects, the HCP “will yield insight into how brain connections underlie brain function, and will open up new lines of inquiry for human neuroscience.”
How do researchers intend to map something so complex and massively parallel as the human brain with its billions upon billions of connections?
To reach its objective, the HCP has to do more than just develop a circuit diagram of the brain. It must also account for the types of connections, the channels, the neurotransmitter levels, the neurotropins, mRNA, and other biochemical and biological variables. NIH intends to collect demographic data as well as data regarding sensory, motor, cognitive, emotional, and the social function for each subject in the HCP study. Additionally, they will be gathering DNA samples and blood (to establish cell lines). They will also fund the development of models to better understand and use these data. Eventually the data might be used to construct human brain circuits — as well as to reconstruct damaged circuits.
In 1953 James Watson and Francis Crick discovered the structure of DNA — the code of instructions for all life on earth. Just 50 years later, in 2003, we were able to view a whole human DNA sequence: a database of some 3 billion letters of genetic code. Developing a comparable connectome database of ourselves and our fellow species such as Tilikum the killer whale may soon shed some light on the wiring of the ultimate connection machine currently known: the human brain.